refactor(csr): use pandas/xarray-style verbs for CSRLinearExpression and Grid methods #1018

Describe the feature you'd like to see

linopy/csr.py names the methods of CSRLinearExpression and Grid as past participles (summed, renamed, reindexed, ...). They return a new frozen object, never mutate. pandas and xarray use plain verbs for the same copy-returning operations (df.rename, da.reindex, fillna, ...), so the CSR layer diverges from the convention of the ecosystem linopy builds on.

Proposal: rename them to pandas/xarray-style verbs. Pure refactor, no behavior change.

# today
expr._csr.aggregated(grid, rows).filled(0.0)
expr._csr.scaled(-1.0)
# proposed
expr._csr.aggregate(grid, rows).fillna(0.0)
expr._csr.mul(-1.0)

Trade-off to decide: LinearExpression already exposes the xarray verbs publicly. After the rename, expr.sum() and expr._csr.sum() share a name but differ in signature and semantics (e.g. the CSR sum takes a grid-level dim list and keeps explicit zero coefficients). The current participles keep the internal sparse layer visibly separate from the public dense API. The _csr attribute and the CSRLinearExpression type arguably mark that boundary well enough on their own.