quantile()/median() don't skip NaN, returning a too-high quantile

[vahid-ahmadi]

MicroSeries.quantile (and therefore median) has no skipna handling. NaN values sort to the end of np.argsort and still contribute their weight to the cumulative distribution, so the inverse-CDF cutoff is pushed upward and the returned quantile is too high:

import numpy as np
from microdf import MicroSeries

MicroSeries([1.0, np.nan, 3.0], weights=[1, 1, 1]).median()   # -> 3.0
MicroSeries([1.0, 3.0], weights=[1, 1]).median()              # -> 1.0

The second line is the reference: dropping the NaN row should give the same answer as never having had it. Instead the NaN row's weight makes up a third of the cumulative distribution, moving the 0.5 cutoff past the second value and onto the largest one.

(Correction to the first version of this issue: I originally wrote "expected 2.0" by comparing with pd.Series.median. That's wrong — pandas interpolates, this method is documented as inverse CDF / survey::svyquantile, which returns an observed value. The correct post-fix answer here is 1.0, and the bug is the NaN weight inflating the CDF, not the absence of interpolation.)

Every neighbouring statistic already takes skipna: mean (#269), count (#292), var/std (#290). quantile/median are the remaining inconsistency, and unlike the others they return a wrong number rather than propagating NaN.

Suggested fix

• Add skipna: bool = True to quantile and median.
• With skipna=True, drop NaN rows (and their weights) before sorting — the same place the existing zero-weight filter from Skip zero-weight rows in MicroSeries.quantile #287 runs.
• With skipna=False, return NaN when any value is NaN, matching mean/var.

Found while reviewing the repo against pandas 3.0.5 / Python 3.13.