refactor(common): clarify coords-entry rules and tighten error labels

[FBumann]

Stacks on top of #732. Small follow-ups from review.

What changes

• Remove dead broadcast_mask — refactor: unify coords-as-truth handling in add_variables/add_constraints #732 claimed it was removed, but the function was still present in linopy/common.py with the old FutureWarning semantics that the rest of the PR explicitly replaced.
• Tuple → list normalization in sequence-form coords. coords=[(0, 1, 2)] previously slipped through _coords_to_dict as a size-only entry but then tripped xarray's (dim_name, values) tuple convention with a confusing error. It now behaves identically to coords=[[0, 1, 2]].
• align_to_coords no longer relabels coords-parsing errors. A TypeError from a bad coords entry (e.g. a stray xarray.DataArray) previously came out as "lower bound could not be aligned to coords: ...", misdirecting users to inspect the bound. It now propagates with its original message; only failures converting value get the labeled wrapper.
• dims= can name an unnamed pd.MultiIndex (parity with pd.Index). The flat .name xarray needs is set on a shallow copy, so the caller's MultiIndex is not mutated. An unnamed MultiIndex with no positional dims= still raises.

Behavior table

This is the spec the docstring of _coords_to_dict and the TestCoordsToDictRules test class both follow (one test per row).

Container forms (top-level coords):

coords is...	Behavior
xarray.Coordinates	Kept dim entries only (MultiIndex level coords dropped).
Mapping	Returned as a shallow dict copy.
sequence of entries	Each entry handled per the rules below.

Sequence-entry rules (i is the position; dims[i] is the matching dims= entry when one exists). An entry is unlabeled if it's an unnamed pd.Index or a bare list / tuple / range / ndarray.

Entry	Naming source	Outcome
pd.Index with .name	.name	accepted
unlabeled entry	dims[i]	accepted
unlabeled entry	— (no dims[i])	skipped — xarray assigns dim_0 etc. downstream
pd.MultiIndex with .name	.name	accepted
pd.MultiIndex without .name	dims[i]	accepted (named on a copy, caller's MI untouched)
pd.MultiIndex without .name	— (no dims[i])	TypeError
anything else (e.g. DataArray)	—	TypeError

Tests

TestCoordsToDictRules in test/test_common.py mirrors the table one-test-per-rule (22 tests, parametrized over list / tuple / range / ndarray for the bare-sequence cases). Each test name is the rule, so the executable spec stays visibly aligned with the docstring.

Test plan

• pytest test/test_common.py::TestCoordsToDictRules — 22 passed
• Full non-solver suite — 1978 passed, 1 skipped
• Doctests on linopy/common.py and linopy/model.py — pass
• Pre-commit (ruff/format/blackdoc/codespell) — clean

🤖 Generated with Claude Code