Clean up / optimize makefile, sandbox builds better.

[Groxx]

• Removed a bunch of cruft from overly-generic makefile copypasta
  (many of it doesn't do anything here, and makes it harder to read).
• Corrected a number of dependency chains. Let me know if I missed
  something! I've tried to be thorough, but it's hard to be sure.
  in any case, now it shouldn't re-do work unless it's necessary, and
  it can even init in parallel (make -j4) from a fresh clone without
  any problems I've seen.
• I was unable to trick Glide into including the necessary packages / versions
  to reliably build things out of vendor... so now there's a versioned_go_build.sh
  script to do that. It'll also warn you if there's a newer version of tool X
  available.
  • Dep can do this pretty easily, add the target command to require... though
    I'm not sure what that implies for users-of-this-lib. Can investigate later.
  • If you change the version vars, it'll automatically build the new version and
    rerun as needed. And they're not removed with make clean, which helps
    keep you from needing to be online to build tools.
• Last but not least: go test ./... is equivalent to (or better than)
  testing each package separately, and is often much faster because it
  doesn't spend time rebuilding everything. And now that we're on Go 1.10, it can
  do coverage across all packages too!

---

Before all changes:

$ make clean
rm -rf cadence-client
rm -Rf ./build
rm -f dummy

$ time make
rm -rf .gen
go get './vendor/go.uber.org/thriftrw'
go get './vendor/go.uber.org/yarpc/encoding/thrift/thriftrw-plugin-yarpc'
thriftrw --plugin=yarpc --pkg-prefix=go.uber.org/cadence/.gen/go/ --out=.gen/go idl/github.com/uber/cadence/cadence.thrift
thriftrw --plugin=yarpc --pkg-prefix=go.uber.org/cadence/.gen/go/ --out=.gen/go idl/github.com/uber/cadence/shared.thrift
go run ./internal/cmd/tools/copyright/licensegen.go --verifyOnly
go build -i -o dummy internal/cmd/dummy/dummy.go
ok  	go.uber.org/cadence/internal	4.175s	coverage: 66.6% of statements
ok  	go.uber.org/cadence/internal/common/backoff	1.046s	coverage: 93.2% of statements
ok  	go.uber.org/cadence/internal/common/cache	1.462s	coverage: 73.1% of statements
ok  	go.uber.org/cadence/internal/common/metrics	1.055s	coverage: 85.0% of statements
ok  	go.uber.org/cadence/internal/common/util	1.024s	coverage: 5.2% of statements
ok  	go.uber.org/cadence/mocks	1.037s	coverage: 6.7% of statements

real	0m20.177s
user	0m11.994s
sys	0m5.099s

$ time make
rm -rf .gen
go get './vendor/go.uber.org/thriftrw'
go get './vendor/go.uber.org/yarpc/encoding/thrift/thriftrw-plugin-yarpc'
thriftrw --plugin=yarpc --pkg-prefix=go.uber.org/cadence/.gen/go/ --out=.gen/go idl/github.com/uber/cadence/cadence.thrift
thriftrw --plugin=yarpc --pkg-prefix=go.uber.org/cadence/.gen/go/ --out=.gen/go idl/github.com/uber/cadence/shared.thrift
go run ./internal/cmd/tools/copyright/licensegen.go --verifyOnly
ok  	go.uber.org/cadence/internal	4.177s	coverage: 66.8% of statements
ok  	go.uber.org/cadence/internal/common/backoff	1.046s	coverage: 93.2% of statements
ok  	go.uber.org/cadence/internal/common/cache	1.460s	coverage: 73.1% of statements
ok  	go.uber.org/cadence/internal/common/metrics	1.051s	coverage: 85.0% of statements
ok  	go.uber.org/cadence/internal/common/util	1.027s	coverage: 5.2% of statements
ok  	go.uber.org/cadence/mocks	1.038s	coverage: 6.7% of statements

real	0m18.736s
user	0m10.713s
sys	0m4.605s

After all changes:

$ make clean
rm -Rf .build
rm -Rf .gen

$ time make
./versioned_go_build.sh go.uber.org/thriftrw v1.11.0 .build/thriftrw
./versioned_go_build.sh go.uber.org/yarpc v1.29.1 encoding/thrift/thriftrw-plugin-yarpc .build/thriftrw-plugin-yarpc
	go.uber.org/yarpc has a newer tag: v1.30.0
thriftrw: idl/github.com/uber/cadence/cadence.thrift
go run ./internal/cmd/tools/copyright/licensegen.go --verifyOnly
touch .build/copyright
go build -i -o .build/dummy internal/cmd/dummy/dummy.go
go test ./... -coverprofile=.build/cover.out -race
?   	go.uber.org/cadence	[no test files]
?   	go.uber.org/cadence/activity	[no test files]
?   	go.uber.org/cadence/client	[no test files]
?   	go.uber.org/cadence/encoded	[no test files]
ok  	go.uber.org/cadence/internal	4.188s	coverage: 66.8% of statements
?   	go.uber.org/cadence/internal/cmd/dummy	[no test files]
?   	go.uber.org/cadence/internal/cmd/tools/copyright	[no test files]
?   	go.uber.org/cadence/internal/common	[no test files]
ok  	go.uber.org/cadence/internal/common/backoff	1.060s	coverage: 93.2% of statements
ok  	go.uber.org/cadence/internal/common/cache	1.490s	coverage: 73.1% of statements
ok  	go.uber.org/cadence/internal/common/metrics	1.065s	coverage: 85.0% of statements
ok  	go.uber.org/cadence/internal/common/util	1.037s	coverage: 5.2% of statements
ok  	go.uber.org/cadence/mocks	1.048s	coverage: 6.7% of statements
?   	go.uber.org/cadence/testsuite	[no test files]
?   	go.uber.org/cadence/worker	[no test files]
?   	go.uber.org/cadence/workflow	[no test files]

real	0m34.329s
user	0m25.309s
sys	0m18.131s

$ time make
go test ./... -coverprofile=.build/cover.out -race
?   	go.uber.org/cadence	[no test files]
?   	go.uber.org/cadence/activity	[no test files]
?   	go.uber.org/cadence/client	[no test files]
?   	go.uber.org/cadence/encoded	[no test files]
ok  	go.uber.org/cadence/internal	4.175s	coverage: 66.8% of statements
?   	go.uber.org/cadence/internal/cmd/dummy	[no test files]
?   	go.uber.org/cadence/internal/cmd/tools/copyright	[no test files]
?   	go.uber.org/cadence/internal/common	[no test files]
ok  	go.uber.org/cadence/internal/common/backoff	1.062s	coverage: 93.2% of statements
ok  	go.uber.org/cadence/internal/common/cache	1.500s	coverage: 73.1% of statements
ok  	go.uber.org/cadence/internal/common/metrics	1.060s	coverage: 85.0% of statements
ok  	go.uber.org/cadence/internal/common/util	1.034s	coverage: 5.2% of statements
ok  	go.uber.org/cadence/mocks	1.042s	coverage: 6.7% of statements
?   	go.uber.org/cadence/testsuite	[no test files]
?   	go.uber.org/cadence/worker	[no test files]
?   	go.uber.org/cadence/workflow	[no test files]

real	0m6.635s
user	0m8.372s
sys	0m2.819s

So the net effect is a slightly slower initial build, and a much faster second+ build.
Seems worth it :)

[Groxx]

... huh. ./versioned_go_build.sh: line 104: DEBUG: unbound variable
bash version difference maybe? can't reproduce. fixed anyway.
and I wonder if grep is old or something on travis, it seems to be interpreting the regex wrong (version tags are being detected as non-versioned). trying a fix and rerunning from what appear to be unrelated test failures.

[coveralls]

Coverage decreased (-0.08%) to 65.462% when pulling 22c013c on Groxx:cleaned into 87a4ffa on uber-go:master.

[Groxx]

I'm thinking of making a bins folder, actually. something that's not cleaned, so versioned-bins don't need to be redownloaded just because someone wanted to rebuild.

[Groxx]

tests are flaky :(

2018-05-18T04:15:25.608Z ERROR internal/internal_task_handlers.go:655 Replay and history mismatch. {"error": "nondeterministic workflow: missing replay decision for ActivityTaskScheduled: (ActivityId:0, Input:[])"}

[yiminc-zz]

Awesome, thanks for doing this!