Handle ambiguity in --min-date and --max-date

[victorlin]

Note

Based on #1913

Description of proposed changes

This fixes a bug where incomplete dates such as --max-date 2018 would not be inclusive, since that had previously resolved to --max-date 2018-01-01.

numeric_date() and get_numerical_date_from_value() are two separate functions that serve similar purposes: converting some date value to a numeric date. The latter has had more recent developments, and crucially will return a date range for ambiguous dates. This is desired for min/max date filters, so I've switched to it.

Related issue(s)

Closes #893

Checklist

• Merge base PR
• Automated checks pass
• Check if you need to add a changelog message
• Check if you need to add tests
• Check if you need to update docs

[victorlin]

Some test matrix jobs are failing at tests/filter/test_relative_dates.py:

tests/filter/test_relative_dates.py:152: in test_filter_relative_dates
    assert output_sorted == output_sorted_expected
E   AssertionError: assert ['SEQ_1', 'SEQ_2', 'SEQ_3'] == ['SEQ_2', 'SEQ_3']
E     
E     At index 0 diff: 'SEQ_1' != 'SEQ_2'
E     Left contains one more item: 'SEQ_3'
E     
E     Full diff:
E       [
E     +     'SEQ_1',
E           'SEQ_2',
E           'SEQ_3',
E       ]
        argparse_params = '--min-date 1D'
…
================== 6 failed, 569 passed, 2 warnings in 14.95s ==================

This needs investigating. To narrow things down a bit, I noticed it happens in test (python=3.10 biopython=1.80 numpy=latest) and not test (python=3.10 biopython=1.80 numpy=1.26.4). I'm unable to reproduce locally with (python=3.13 biopython=1.85 numpy=2.3.2), so it's not just the numpy version.

(unrelated to this file, just needed to start a thread)