`test(features): enforce lexicon values for water-level descriptor fields

[marissafichera]

Why

This PR addresses the following problem / context:

• The water-level-csv.feature still allowed free-text behavior where sample_method, level_status, and data_quality should be enum/lexicon-constrained.
• We need feature coverage that explicitly requires lexicon-valid values for those three fields.

How

Implementation summary - the following was changed / added / removed:

• Updated tests/features/water-level-csv.feature only.
• Added a happy-path requirement that, when provided, sample_method, level_status, and data_quality must be valid lexicon values.
• Replaced the free-text acceptance scenario with a negative validation scenario that fails upload when those fields contain invalid lexicon values.
• Kept expectations consistent with existing validation style:
  • non-zero exit code
  • validation errors identify row/field
  • no rows imported

Notes

Any special considerations, workarounds, or follow-up work to note?

• This is a spec-only PR (no Python implementation changes).
• If current step definitions still model the old free-text behavior, they may need a small follow-up to match this updated feature.

[Copilot]

Pull request overview

Updates the water-level CSV bulk-upload BDD spec to require lexicon-constrained values for water-level descriptor fields, replacing prior free-text acceptance expectations.

Changes:

• Added a happy-path requirement stating sample_method, level_status, and data_quality must be lexicon-valid when provided.
• Replaced the free-text acceptance scenario with a negative validation scenario for invalid lexicon values.

[ksmuczynski]

Looks great, approved!