feat(revsets): add merges() and allow branches() to filter branch names

[claytonrcarter]

• I have found merges() to be useful, in particular on public commits.
• for branches(<pattern>) see Tell `sync` to ignore certain branches #1117
• branches(<pattern>) needs more work; it's currently mush slower that I would expect, and it seems to not match some branchs (at least it didn't match all the branches I expected when I used it to push these recent PRs) Edit: I believe that this is fixed.

[claytonrcarter]

branches() needs more work; it's currently mush slower that I would expect, and it seems to not match some branchs (at least it didn't match all the branches I expected when I used it to push these recent PRs)

For example, on my copy of the git-branchless repo, I have approx 1600 commits and 7 branches, as follows:

⋮
◆ 5871499 2h (ᐅ master) Change 'suggest' to 'discuss'
┣━┓
┃ ◯ 2bd4827 5s (crc/submit-dry-run) feat(submit): add --dry-run flag
┣━┓
┃ ◯ 306e9b4 5s (crc/fix-status-spaces, local) fix(status): rudimentary support for filenames with spaces
┣━┓
┃ ◯ 9661224 5s (crc/sync-resolve-stack) feat(sync): resolve revset arguments to their entire stacks
┣━┓
┃ ⊘ 2 omitted commits
┃ ⋮
┃ ◯ 620a061 5s (move-fixup-on-disk) temp(on disk fixup): Try to fix abandoned branches
⋮
⊘ 1 omitted commit
⋮
◯ fa5342c 5s (crc/revsets) feat(revset): allow branches() to accept a text pattern

Running git sl 'branches(crc)' takes around 3s to run, and seems to process every commit in the repo. Also, git submit 'branches(crc)' --dry-run doesn't match/report the crc/fix-status-spaces branch, which I'm currently assuming has something to do w/ the other branch that points to that commit. (Hmm, on second thought, perhaps that's an issue w/ the implementation of #1129?) Edit: nope, that was a separate issue #1131

[claytonrcarter]

Running git sl 'branches(crc)' takes around 3s to run, and seems to process every commit in the repo.

This was my misunderstanding of how revset Matchers work. I hadn't connected that they work on all visible commits, and there is currently no way to restrict that. (This is the first revset feature – that I can tell – that would make any sense to operate on anything other than all visible commits, because we know that branches are (probably) a subset of visible commits.)

[claytonrcarter]

Sorry, I lost track of this. As I recall, I have addressed the review as best I could figure out; see my comments from 11/13. I think I just forgot to un-draft this after doing so. Thank you!

[arxanas]

Looks good pending a few comments!

[claytonrcarter]

Looks good pending a few comments!

Great, thank you!