extract: Warn when skipping an f-string with substitutions

[ColinHouse]

_(f"Hello {name}") cannot be translated, so extraction skips it. Since #915
replaced eval with ast, it is skipped silently: pybabel extract exits 0,
writes no message, and prints nothing. The string never reaches translators and
ships untranslated, with no signal to the author.

This warns instead, in the same format as the existing "Empty msgid" warning:

app.py:2: warning: Skipping f-string with substitutions: it cannot be
translated as written.  Extract a format string instead,
e.g. _("Hello {name}").format(name=name).

There are two code paths, because f-string tokenisation changed in 3.12:

• before 3.12 the whole f-string is one STRING token and _parse_python_string
  returns None;
• from 3.12 (PEP 701) a substitution arrives as a non-FSTRING_* token and the
  f-string is dropped where current_fstring_start is reset.

Both are covered, and the suite was run on 3.11 and 3.12.

It warns rather than raising, as the issue body asks ("warn the user about the
incorrect usage instead of simply crashing"). Raising would break builds that
extract successfully today. Happy to make it an error if you prefer.

A constant f-string such as _(f"spam" f"eggs") is still extracted and does not
warn; an f-string outside a translation call is untouched.

The stale # TODO: we could raise an error or warning when not all nodes are constants is replaced, since this implements it.

fixes #715

[akx]

LGTM, thanks!

[codspeed]

Merging this PR will not alter performance

✅ 88 untouched benchmarks

---

_{Comparing ColinHouse:warn-skipped-fstring (6f30522) with master (6ba6701)}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.38%. Comparing base (6ba6701) to head (6f30522).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1343      +/-   ##
==========================================
+ Coverage   93.37%   93.38%   +0.01%     
==========================================
  Files          27       27              
  Lines        4919     4927       +8     
==========================================
+ Hits         4593     4601       +8     
  Misses        326      326              

Flag	Coverage Δ
macos-14-3.10	92.40% <62.50%> (-0.05%)	⬇️
macos-14-3.11	92.34% <62.50%> (-0.05%)	⬇️
macos-14-3.12	92.59% <87.50%> (-0.01%)	⬇️
macos-14-3.13	92.59% <87.50%> (-0.01%)	⬇️
macos-14-3.14	92.57% <87.50%> (-0.01%)	⬇️
macos-14-3.15	92.57% <87.50%> (-0.01%)	⬇️
macos-14-3.8	92.28% <62.50%> (-0.05%)	⬇️
macos-14-3.9	92.34% <62.50%> (-0.05%)	⬇️
macos-14-pypy3.10	92.40% <62.50%> (-0.05%)	⬇️
ubuntu-24.04-3.10	92.42% <62.50%> (-0.05%)	⬇️
ubuntu-24.04-3.11	92.36% <62.50%> (-0.05%)	⬇️
ubuntu-24.04-3.12	92.61% <87.50%> (-0.01%)	⬇️
ubuntu-24.04-3.13	92.61% <87.50%> (-0.01%)	⬇️
ubuntu-24.04-3.14	92.59% <87.50%> (-0.01%)	⬇️
ubuntu-24.04-3.15	92.59% <87.50%> (-0.01%)	⬇️
ubuntu-24.04-3.8	92.30% <62.50%> (-0.05%)	⬇️
ubuntu-24.04-3.9	92.36% <62.50%> (-0.05%)	⬇️
ubuntu-24.04-pypy3.10	92.42% <62.50%> (-0.05%)	⬇️
windows-2022-3.10	92.42% <62.50%> (-0.05%)	⬇️
windows-2022-3.11	92.36% <62.50%> (-0.05%)	⬇️
windows-2022-3.12	92.60% <87.50%> (-0.01%)	⬇️
windows-2022-3.13	92.60% <87.50%> (-0.01%)	⬇️
windows-2022-3.14	92.58% <87.50%> (-0.01%)	⬇️
windows-2022-3.15	92.58% <87.50%> (-0.01%)	⬇️
windows-2022-3.8	92.39% <62.50%> (-0.05%)	⬇️
windows-2022-3.9	92.35% <62.50%> (-0.05%)	⬇️
windows-2022-pypy3.10	92.42% <62.50%> (-0.05%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[ColinHouse]

Sorry — this closed itself because I deleted my fork by accident, minutes after your review. Nothing to do with the review.

Reopened as #1350 with the same commit (6f30522) and a byte-identical diff.