bug: flat comparison edge-cases causing mismatches / pipeline failures

[ThomasHepworth]

Summary

Several flat/unit comparison edge-cases aren’t parsed/handled, leading to unexpected non-matches and occasional pipeline failures.

Issues / examples

1. BASEMENT vs FLAT A vs FLAT 1

   • Comparisons unsupported.
   • Examples: BASEMENT ↔ FLAT A, BASEMENT ↔ FLAT 1, FLAT A ↔ FLAT 1

2. Floor ordinals not parsed

   • Can’t parse: FIRST AND SECOND FLOORS. Should ideally result in an array or equivalent.

3. Range handling

   • Example: FLAT 2 92 ↔ FLAT 2 92-92A (range) doesn’t behave as expected. We parse the letter in the range as a flat identifier, when it's clearly Flat 2 in house numbers 92 to 92A.

4. Duplicate numeric token count ignored

   • Example: 3, 3 LOVE LANE ↔ 3 LOVE LANE (extra numeric token not accounted for)

[ThomasHepworth]

Multi-floors code could do with some refinement:
https://github.com/moj-analytical-services/uk_address_matcher/blob/main/uk_address_matcher/cleaning/steps/token_parsing.py#L158