Make (*worker).fetchJob() x3 faster

[vadiminshakov]

Performance of main worker loop improved (x3). Added interfaces to make worker testable. Old implementations not deleted, just for comparison, see benchmark added.

go test -bench=Benchmark_fetchJob -run=xxx . -cpuprofile=profile.pprof -benchtime=20s

goos: darwin
goarch: amd64
pkg: github.com/sbermarket-tech/work
cpu: Intel(R) Core(TM) i7-9750H CPU @ 2.60GHz
Benchmark_fetchJob/old_fetcher-12                  23812            938909 ns/op
Benchmark_fetchJob/new_fetcher-12                  79836            301719 ns/op

[vadiminshakov]

@aermolaev

[vadiminshakov]

I deleted all the staff you worry about (nonstd marshaller, bench, interfaces, stubs)