REGR: perf regression in Series.combine_first

[lukemanley]

• closes BUG: combine_first major performance regression #55845
• Tests added and passed if fixing a bug or adding a new feature
• All code checks passed.
• Added type annotations to new arguments/methods/functions.
• Added an entry in the latest doc/source/whatsnew/v2.2.1.rst file if fixing a bug or adding a new feature.

import pandas as pd
import numpy as np

N = 1_000_000


# different indexes
s1 = pd.Series(np.random.randn(N))
s1 = s1[::2].sample(frac=1.0)
s2 = pd.Series(np.random.randn(N))
%timeit s1.combine_first(s2)

# 523 ms ± 7.4 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)   <- main
# 227 ms ± 4.33 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)  <- PR


# same index
s1 = pd.Series(np.random.randn(N))
s1[::2] = np.nan
s2 = pd.Series(np.random.randn(N))
%timeit s1.combine_first(s2)

# 282 ms ± 5.64 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)   <- main
# 7 ms ± 64.3 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)  <- PR

[WillAyd]

Surprising to see this here. Maybe more motivation for #56518

[lukemanley]

yes, a bit unfortunate. Alternatively, the sparse issue was previously xfailed and could be xfailed again in order to avoid this:

https://github.com/pandas-dev/pandas/pull/51777/files#diff-16458c6ca46c7c2e895b1fad1300009d8285d19cdbef3023a33f520d15bc7085L333

[mroeschke]

Thanks @lukemanley