Always store MicroDataFrame weights as a Series

[vahid-ahmadi]

Closes #303

Problem

Two methods assigned a bare np.ndarray to self.weights, while the rest of the class assumes it is an index-aligned pd.Series:

• nullify_weights — self.weights = np.ones(len(self))
• set_weight_col (deprecated) — self.weights = np.array(self[column])

d = MicroDataFrame(pd.DataFrame({"x": [1, 2, 3]}), weights=[4, 5, 6])
d.nullify_weights()
d.equals(d)   # AttributeError: 'numpy.ndarray' object has no attribute 'equals'

__getitem__ and _MicroLocIndexer.__getitem__ both call self.weights.reindex(...), so column and row selection on a nullified frame were one step from the same failure. set_weight_col additionally skipped the length validation and index alignment that set_weights performs.

Change

Both methods now delegate to set_weights, which is the single place that validates length and builds pd.Series(..., index=self.index, dtype=float).

One intentional behaviour change: nullify_weights now clears weights_col (via set_weights), which is correct — after nullifying, the weights no longer come from a column. preserve_old handling in set_weight_col is unchanged; it still runs before the delegation.

Verification

• Two new tests; both fail on main (AttributeError on equals) and pass with the change
• make test — 64 passed on pandas 3.0.5 and pandas 2.3.3

Note: the lint job will be red until #307 merges (unrelated ruff 0.16 issue on main).