Refactor branch list using src-d/go-git and add GetRefs function

[lafriks]

GetRefs will be needed to implement Refs API

Add src-d/go-git dependency and refactored GetBranches to use this lib.

Old using git command:
BenchmarkRepository_GetBranches-4           1000           1505930 ns/op
New using src-d/go-git:
BenchmarkRepository_GetBranches-4          10000            150774 ns/op

[codecov-io]

Codecov Report

Merging #133 into master will decrease coverage by 0.21%.
The diff coverage is 63.63%.

@@            Coverage Diff             @@
##           master     #133      +/-   ##
==========================================
- Coverage   33.54%   33.33%   -0.22%     
==========================================
  Files          25       28       +3     
  Lines        1407     1761     +354     
==========================================
+ Hits          472      587     +115     
- Misses        862     1096     +234     
- Partials       73       78       +5

Impacted Files	Coverage Δ
ref.go	0% <0%> (ø)
repo_branch.go	12.32% <58.33%> (+12.32%)	⬆️
repo_ref.go	70% <70%> (ø)
tree.go	68.42% <0%> (-6.58%)	⬇️
signature.go	64% <0%> (-6%)	⬇️
repo_tree.go	35.71% <0%> (-5.96%)	⬇️
tree_blob.go	40.54% <0%> (-4.63%)	⬇️
commit_info.go	63.92% <0%> (-4.08%)	⬇️
git.go	27.27% <0%> (-3.29%)	⬇️
parse.go	57.62% <0%> (-3.16%)	⬇️
... and 18 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f3afe30...0b52498. Read the comment docs.

[lunny]

@lafriks the benchmark is for a bare git repository?

[lafriks]

@lunny yes

[lunny]

It seems we only have a bare repo on testdata

[lafriks]

@lunny is that a problem?

[lunny]

I prefer pure go library than git command. But for performance measure, I think a bare repo maybe not a good example.

[lafriks]

We store all git repos as bare server-side so that is measurement most important for gitea case