cmd/gofmt: align renamed imports

[dsnet]

Currently, my import block looks something like:

import (
	"bytes"
	"math"
	"reflect"
	"testing"

	"github.com/google/go-cmp/cmp"
	"github.com/google/go-cmp/cmp/cmpopts"
	"google.golang.org/protobuf/internal/encoding/pack"
	pimpl "google.golang.org/protobuf/internal/impl"
	"google.golang.org/protobuf/internal/pragma"
	ptype "google.golang.org/protobuf/internal/prototype"
	"google.golang.org/protobuf/internal/scalar"
	pref "google.golang.org/protobuf/reflect/protoreflect"
	preg "google.golang.org/protobuf/reflect/protoregistry"
	piface "google.golang.org/protobuf/runtime/protoiface"

	proto2_20180125 "google.golang.org/protobuf/internal/testprotos/legacy/proto2.v1.0.0-20180125-92554152"
)

Most of the import paths have a shared prefix, but the lack of aligning makes this hard to see. Preferably, gofmt would align the imports like:

import (
	"bytes"
	"math"
	"reflect"
	"testing"

	       "github.com/google/go-cmp/cmp"
	       "github.com/google/go-cmp/cmp/cmpopts"
	       "google.golang.org/protobuf/internal/encoding/pack"
	pimpl  "google.golang.org/protobuf/internal/impl"
	       "google.golang.org/protobuf/internal/pragma"
	ptype  "google.golang.org/protobuf/internal/prototype"
	       "google.golang.org/protobuf/internal/scalar"
	pref   "google.golang.org/protobuf/reflect/protoreflect"
	preg   "google.golang.org/protobuf/reflect/protoregistry"
	piface "google.golang.org/protobuf/runtime/protoiface"

	proto2_20180125 "google.golang.org/protobuf/internal/testprotos/legacy/proto2.v1.0.0-20180125-92554152"
)

[bcmills]

CC @griesemer

[griesemer]

I'm ok with this if we decide to make this change. Marking for Go 1.14, pending decision.

[dsnet]

I'm doing some LSCs (large-scale changes) and I'm observing that a frequent cause of merge conflicts is due to alignment changing many more lines than the change was strictly about (I commonly run into this issue with Go map literals and trailing comments).

I should note that while alignment brings a positive benefit to readability, it brings the detriment of higher probability of a merge conflict when two authors modify the same file and add/remove some imports.

[myitcv]

Just noting in passing that this was discussed in #tools on Slack:

https://gophers.slack.com/archives/C0VPK4Z5E/p1580142385130500

@josharian noted in his response that he had previously looked at this in the form of https://github.com/whatsnu/semdiffstat.

I never got round to looking at how/whether it was possible to use something like semdiffstat during a merge/rebase (which was my main use case, much like yours).