Fix IndexError in DateTimeParser.parse on malformed format strings

[aldorizona10-glitch]

Fixes #1191.

Problem

A malformed format string can leave a token in fmt_tokens without a corresponding capture group in the compiled pattern (the bracket/escape reinsertion in _generate_pattern_re can desync the two). parse() then called match.group(token), which raised a raw IndexError: no such group instead of Arrow's documented ParserMatchError.

Minimal reproducer (the report's fuzzed PoC reduces to this):

from arrow.parser import DateTimeParser
DateTimeParser().parse("foo", "[|(\\]s")   # IndexError: no such group

Fix

Guard the group lookup in parse() so a missing group raises ParserMatchError — matching the handling already used a few lines below when a matched group's value is None. parse() already documents :raises ParserMatchError:, so this makes the behaviour consistent with the contract.

Testing

Ran the parser suite locally: 724 passed, plus a new regression test (test_parse_malformed_fmt_no_match_group). black, flake8, and isort are all clean.

---

Implemented with AI assistance (Claude Code, Opus 4.8); I reviewed the change and verified it locally as described above. Happy to add a CHANGELOG.rst entry referencing this PR once it has a number.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (2224255) to head (0981b88).

Additional details and impacted files

@@            Coverage Diff            @@
##            master     #1335   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           10        10           
  Lines         2315      2318    +3     
  Branches       358       358           
=========================================
+ Hits          2315      2318    +3     

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

[aldorizona10-glitch]

Heads-up for reviewers: the one red check — windows-latest / pypy-3.11 — failed during environment setup, not in the tests. pip hit a transient network error building pyyaml's build dependencies (Connection broken: IncompleteRead), so tox never reached the suite. All other 26 checks pass, including ubuntu-latest / pypy-3.11 and codecov (100%). A re-run should clear it — happy to push a trivial change to re-trigger if that's easier.

[aldorizona10-glitch]

Correction to my earlier note: a re-run did not clear it — and it turns out windows-latest / pypy-3.11 fails the same way on master and on the other current open PRs (#1330–#1334), all with pip failing to build pyyaml's dependencies on that runner. So it's a pre-existing, environment-wide CI issue rather than anything specific to this change, which is green on all 26 other checks (codecov 100%). Flagging in case the required-check config needs a look on your side.