Release 0.22rc2

[adrinjalali]

This bumps the version for the 0.22 release to RC2, and includes [almost] all the PRs merged in the meantime.

Please don't squash when merging.

the rebase log:

pick 593d3de124 DOC add min version for conda and example of flag set by conda (#15619)
pick bbd1964c53 DOC Fix 'Title underline too short' sphinx warning. (#15620)
pick 859953023f DOC CountVectorizer documentation update (#15498)
pick 25e72d341e DOC fix early stopping doc in HGBT (#15623)
pick e650a207ef ENH Adds plot_confusion matrix (#15083)
pick f1d883252b DOC Adds note regarding plot partial dependence with the same axes (#15599)
pick 343214085a ENH KernelPCA raises error in case of numerical/conditioning issues (#12145)
pick 2885a06c94 BUG Fixes partial dependence bug by setting axes to be invisible (#15572)
pick 7042019c8d BLD Prevent hyphenation in parameter description (#15499)
pick 97958c174f FIX fix bug when warm starting with early stopping in Hist GBDT (#15624)
pick 080d0e5595 ENH add normalize parameter to confusion_matrix (#15625)
pick d5bd4e6a58 DOC mention other plotting utilities in highlights (#15569)
pick 3e4ad0567d MNT Fix whats new link on the home page (#15634)
drop 947f37e3a1 MNT bump version to 0.23.dev0 and add new whats_new (#15631)
pick 004426ab53 ENH Consistent checks for sample weights in linear models (#15530)
pick 25a88b4a2b DOC docstrings validation in TfidfVectorizer (#15509)
pick ff878d60aa DOC add versionadded to OneHotEcoder, LabelEncoder (#15506)
pick 8360786895 ENH Allowing Gaussian process kernels on structured data (#15557)
pick cd81de1ccc MAINT Improve method detection in numpydoc validation script (#15532)
pick d85d1e9ab1 ENH Sets xlabel only if it is not defined in PDP (#15610)
pick 5483b0ecd1 DOC Use anonymous reference in readme for rst standard to be happy (#15643)
pick ee09095cea DOC minor fixes in whats_new for the release (#15632)
pick d84785c5e1 [MRG] ENH: Add scaling to convergence warning for LBFGS (#15571)
pick 93661c7fdd included default values in SparseCoder class doc (#15600)
pick 54e78b7b07 DOC Birch numpydoc docstring validation (#15511)
pick 98cb91b545 DOC numpydoc validation for VotingClassifier (#15497)
pick c311efd501 FIX accept pandas Index for feature_names in PDP (#15639)

@jnothman this looks okay?
also ping @NicolasHug

[glemaitre]

As candidates to be included:

• MNT Reorder parameters in plot_confusion_matrix #15645
• ENH do not allocate local arrays in Ridge*CV of store_cv_values is False #15652

[jnothman]

Looks fine if all we're doing is effectively branching later and there's nothing controversial in master. Let's do it: announce it before the end of the week

[jnothman]

I mean ideally the rc1 would have been associated with a feature freeze after which you only pick fixes, especially regressions and changes to new features.

[adrinjalali]

If I understand correctly, this should be a rebase and merge, not a squash and merge, but we have it disabled on the repo. Am I missing something?

[jnothman]

I've previously just done it on the command-line (or we could enable it in the repo).

git checkout 0.22.X
git merge --ff-only release-0.22rc2
git push upstream

or something

[jnothman]

I've now permitted Rebase and Merge as it is occasionally useful (and safer)