refactor(farm_base): extract farm.rank.mixin from observation + fence

[dnplkndll]

What changed

Both farm_observation.urgency_rank and farm_fence.condition_rank implemented the same pattern: a Selection field whose alphabetical order doesn't match priority order, paired with a stored-computed Integer mirror used as the sort key in _order.

This PR extracts that mechanism into farm.rank.mixin (in farm_base). Subclasses now declare just:

class FarmObservation(models.Model):
    _inherit = ["mail.thread", "farm.rank.mixin"]
    _order = "rank desc, observation_date desc"
    _rank_selection_field = "urgency"
    _rank_value_map = {"low": 0, "med": 1, "high": 2}

…and pick up the stored, indexed rank integer the mixin contributes. The two consumers each drop ~10 lines.

Migration

Renaming urgency_rank → rank and condition_rank → rank is a column rename. farm-pack just shipped these models at 19.0.1.0.0 and no production installs exist yet, so this PR bumps to 19.0.1.1.0 and accepts the rename without a migration script.

Tests

• farm_base/tests/test_rank_mixin.py — new — confirms the abstract model registers and the rank field declares store=True, index=True, readonly=True.
• Existing regression tests in farm_observation + farm_fence (test_urgency_sort_priority_not_alphabetical, test_condition_sort_priority_not_alphabetical) keep passing — they assert search-result order, which doesn't care which field holds the rank.

Test plan

• CI green on 19.0
• Smoke install on local doodba: install farm_observation + farm_fence, confirm the rank column is created and grouped kanbans put high-urgency / repair items on top