Annotate merge-base comparisons; improve report links

[IanButterworth]

Comments in the code indicated we use the merge base, but it doesn't actually seem we do.
Actually, it's just a footgun of using the vs=:master form when you just want a PR diff. Adds user guidnace.

Developed with claude:

Updated

---

Add transparency around merge-base comparisons (introduced in #134):
reports and reply comments now indicate when a merge-base was used and
show which branch it was computed from. When vs explicitly references
the PR's base branch (e.g. vs=":master" on a PR targeting master),
add a hint suggesting to omit vs to get the merge-base comparison.

Also:

• Show SHA range in Comparison Diff/Range links instead of generic "link"
• Remove unused prnumber parameter from build_julia!
• Remove stale comments about merge commit building (never implemented)
• Simplify build_benchmarksjulia! call site

[maleadt]

We are using the merge base, but only when vs is unspecified: #134
I'm don't think we want the proposed behavior. When specifying vs, to me that indicates you really want to compare against that branch or commit, and not against some derivative point in history.

[IanButterworth]

Ah ok. I remember that now. I think the current common runbenchmarks(ALL, vs=":master") has become a bit of a footgun. The report/comment should probably be clearer what's happening/guide the user.

[IanButterworth]

Updated to just improve user guidance

[maleadt]

The recommendation should just be to do runbenchmarks(ALL) or runtests(), both of which use the merge base, while allowing people (mostly @KristofferC) to compare against specific branches using vs.

[IanButterworth]

Yeah, I'm just seeing a lot of use of vs=:master still https://github.com/search?q=repo%3AJuliaLang%2Fjulia+%22runbenchmarks%22&type=pullrequests&s=created&o=desc

This just tries to make it clearer.

[maleadt]

With runbenchmarks that was also what people had to do until recently (presumably that's fixed now) because the benchmark worker hadn't been updated in a long time, and didn't include the above PR.

[IanButterworth]

Ah I see.