PERF: Use generator expression for Blocks.replace_list

[lithomas1]

• closes #xxxx (Replace xxxx with the GitHub issue number)
• 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/vX.X.X.rst file if fixing a bug or adding a new feature.

       before           after         ratio
     [852518eb]       [3902ad1a]
     <maybe-convert-less-mem^2>       <replace-use-less-mem>
-      2.68±0.02s       2.35±0.03s     0.88  series_methods.Replace.time_replace_dict(1000)
-       2.73±0.1s       2.38±0.04s     0.87  series_methods.Replace.time_replace_list(1000)
-            203M             105M     0.52  series_methods.Replace.peakmem_replace_dict(100)
-            203M             105M     0.52  series_methods.Replace.peakmem_replace_list(100)
-           1.11G             106M     0.10  series_methods.Replace.peakmem_replace_list(1000)
-           1.11G             106M     0.10  series_methods.Replace.peakmem_replace_dict(1000)

SOME BENCHMARKS HAVE CHANGED SIGNIFICANTLY.
PERFORMANCE INCREASED.

inplace=True is not covered by benchmarks but should be unchanged.

[jbrockmendel]

is there a chance of an API change here (not necessarily a bad one) with something like

src_list = [1, 2, 3]
dest_list = [2, 3, 4]
values = np.array([1, 2, 3])

IIUC in the status quo we end up with [2, 3, 4] but with this we'd end up with all-4s?

[lithomas1]

Ah, looks like this would be invalid for inplace=True. Good catch!

I'll also add a test case, since it looks like this isn't caught by CI.

[jbrockmendel]

Might address #6697?

[lithomas1]

Might address #6697?

Only partially. This optimization only works for inplace=False. We'd need to write our own replace function, to avoid generating the masks.

[jbrockmendel]

Needs rebase, otherwise LGTM

[lithomas1]

@meeseeksdev backport 2.x

[lithomas1]

@meeseeksdev backport 2.0.x