Parse mixed-case and compound named clock values

[OskarEichler]

Summary

The named-minute alternation tries shorter words before compound forms, so twenty consumes the prefix of twenty-one. It also recognizes case-insensitive words but looks them up without lowercasing. Try longer minute names first and normalize the captured clock words.

Reproduction

Run after require 'fugit':

p Fugit.parse_nat('every day at five twenty-one').to_cron_s # "21 5 * * *"
p Fugit.parse_nat('every day at FIVE TEN').to_cron_s # "10 5 * * *"
# both previously nil

Verification

• rbenv Ruby 4.0.6, TZ=UTC. Master baseline and this individual patch: 1,524 existing tests, 3,379 assertions, zero failures, one existing pending test.
• 6 focused checks for this change, run outside the repository. The combined release-based branch passes 1,517 existing RSpec examples, zero failures, one pending, plus 998 focused checks.
• Also checked the combined master changes with unmodified registry EtOrbi 1.4.1/Raabro 1.5.0: 1,524 tests/3,379 assertions, zero failures, one pending, and 997 focused checks. The extra blank-line check requires the independently fixed Raabro empty-input behavior; it is not claimed to pass on unmodified Raabro.
• Individual/release checks used the same versions of EtOrbi/Raabro with their separately proposed fixes. Syntax and whitespace checks pass. Combined source has the same 18 pre-existing Lint warnings, with no new warnings. Release gem packaging checked.

Compatibility and limitations

Previously rejected valid compound/mixed-case clock names are accepted. Existing lowercase names and am/pm behavior remain unchanged.

No dependency/version upgrades, test additions or test modifications are included. The task's no-new-tests constraint was followed; reproduction/check scripts stayed outside the repository. Other Ruby versions and platforms await upstream CI. No production execution was performed. Prepared with Codex assistance.