🐛 sphinx-needs: an empty need.FIELD selector fails the call, as an unset one does

[chrisjsewell]

What

Package: packages/sphinx-needs.

A need.<field> argument that selects what a built-in reads (copy's id, field and filter; calc_sum's field,
filter and links_only; check_linked_values' field and filter; links_from_content's id and filter) now fails the
call when its field holds "", exactly as it already did when the field holds None:

.. req:: Uses a selector its need does not set
   :id: E_COPY
   :summary: [[copy("title", need.src)]]

With src an untyped extra option (needs_extra_options = ["src"]) that the need does not set, src is "", and the
call ran as copy("title", ""), which copies the need's own title, silently. It is now one
needs.dynamic_function warning, Error while applying need to function 'copy': need.src selects what the call reads, and is not set, and the field keeps its placeholder (empty here). The same holds for the call in an :ndf: role or a
needtable :style_row: (they share the check), and the dependency order reads nothing for such a call. A need.<field>
in a value argument (check_linked_values' result or search_value, copy's upper) still passes "" on.

• functions/order.py: one _UNSET = (None, "") used by unset_selector() (the run-time check of the field, the role
  and :style_row:) and by the order's own reading of a selector.
• Docs: dynamic_functions.rst's need.<field> paragraph says "unset or empty" (and that a value argument is None,
  or "" for an untyped extra option, when unset); the 9.0.0 entry of the changelog likewise, in the Unreleased
  entry "Dynamic functions and variants are computed in dependency order", whose cite list gains this pull request.

Why

#2064 (derived values), #2134 (phase 1, merged), which made an unset (None) selector fail the call instead of
letting copy fall back to the need's own field. ubCode's review of its own phase 1 found that an untyped extra option
left unset holds "", not None, so the same silent fallback remained for the most common kind of extra option, and
links_only=need.src summed every need instead of the need's links. Both tools now treat "" in a selector as unset,
with the same finding.

Tests

• test_an_empty_need_attribute_selector_fails_the_call (tests/test_dynamic_functions_strata.py): an untyped extra
  option left unset in copy's id, calc_sum's field, check_linked_values' field and links_only, each None and
  one needs.dynamic_function naming need.src; the :ndf: role renders ??; upper=need.src is a value and copies
  normally. Red before the change: {'E_COPY': "copy's need"} != {'E_COPY': None} (the own-field copy) and
  {'E_ONLY': 2.0} != {'E_ONLY': None} (every need summed).
• tests/test_functions_order.py: two rows of the edge table, an empty selector of copy and an empty links_only,
  read nothing. Red before: {'deps': [('RD', 'summary')]} != {'deps': []} and
  {'columns': [(Column(field='hours', candidates=None), None)]} != {'columns': []}.

The full suite (uv run poe test-needs -n 4 tests/) passes but for the 50 graphviz renders that need dot (the same 50
on master on a machine without it).

Checklist

• I wrote this change myself and have read every line of it; it was not generated automatically from an issue.
  Built in an orchestrated session from ubCode's review finding and the code; the tests were written first and shown
  red, and every line was read.
• I ran the package's tests (uv run poe test-needs) and they pass (the graphviz renders excepted on a machine
  without dot, identically to master).
• Documentation is updated where behaviour or options change (docs/dynamic_functions.rst).
• The package's docs/changelog.rst has the change, in the existing Unreleased derived-values entry (its
  :pr: cite now names this number).
• uv run poe lint and uv run poe typecheck pass.

Refs #2064.

When squashing, strip the Co-authored-by trailer GitHub proposes.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.78%. Comparing base (1bba2ea) to head (4aa3756).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #2136   +/-   ##
=======================================
  Coverage   92.78%   92.78%           
=======================================
  Files         137      137           
  Lines       20158    20159    +1     
=======================================
+ Hits        18704    18705    +1     
  Misses       1454     1454           

Flag	Coverage Δ
ai-index	100.00% <ø> (ø)
codelinks	95.35% <ø> (ø)
mounts	94.27% <ø> (ø)
pytests	92.34% <100.00%> (+<0.01%)	⬆️
reports	88.82% <ø> (ø)
ub-test-reports	90.20% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.