Fix ValueError when addplot column is all NaN

[wavebyrd]

Summary

When all values in an addplot column are NaN, _addplot_columns filters
out NaN values into yd, which becomes an empty list. Calling
np.nanmax() / np.nanmin() on that empty list raises:

ValueError: zero-size array to reduction operation maximum which has no identity

This adds a length check before the nanmax/nanmin calls. When there is
no real data, secondary_y auto-detection is skipped and the column is
plotted on the primary y-axis. matplotlib already handles all-NaN series
gracefully in the actual plot/scatter/bar calls, so nothing else needs
to change.

Reproduction

import numpy as np
import pandas as pd
import mplfinance as mpf

# (generate OHLC DataFrame `df`)
df["signal"] = np.nan
ap = mpf.make_addplot(df["signal"])
mpf.plot(df, type="candle", addplot=ap)  # ValueError before fix

Test plan

• Verified the reproduction script from Bug Report: Errors occur for NaN columns. #672 runs without error
• Verified normal addplot data (non-NaN) still works
• Verified partial NaN addplot data still works
• Verified multiple addplots with one all-NaN column still works

Fixes #672

[rcomer]

This account has been blocked from the Matplotlib organisation for automating PRs matplotlib/matplotlib#31296 (comment)