PERF-#6762: Carry dtypes information in lazy indices

[dchigarev]

What do these changes do?

This PR adds ._dtypes property to ModinIndex, one can pass it to the object's constructor and indicate dtypes of index levels, even if the index values are yet unknown.

This was done mainly to preserve dtypes in the following case:

df.dtypes # known_dtypes: {"a": int, ...}
res = df.groupby("a", as_index=True).sum()
res # 'a' column is now in the index, but we lost its dtypes
res = res.reset_index(drop=False)
res.dtypes # cols_with_unknown_dtypes: ["a"]

With the changes in this PR, the presented scenario will preserve dtype for "a" column in the end.

• first commit message and PR title follow format outlined here

  NOTE: If you edit the PR title to match this format, you need to add another commit (even if it's empty) or amend your last commit for the CI job that checks the PR title to pick up the new PR title.

• passes flake8 modin/ asv_bench/benchmarks scripts/doc_checker.py
• passes black --check modin/ asv_bench/benchmarks scripts/doc_checker.py
• signed commit with git commit -s
• Resolves Support carrying dtypes in an index cache #6762
• tests added and passing
• module layout described at docs/development/architecture.rst is up-to-date

[anmyachev]

Why such restrictions?

[dchigarev]

These restrictions describe a simple case when 'by' are columns from the same dataframe. I don't want to handle other cases in this PR as pandas is always tricky in how it handles groupby with mixed by

[dchigarev]

added an in-code comment about this

[anmyachev]

LGTM!