feat(rust): POC metadata reading and writing

[coastalwhite]

This PR illustrates a basic workflow for keeping track of and utilizing metadata. Currently, this is still all behind POLARS_METADATA_USE=experimental, but now we can start implementing a lot of places where metadata can be inferred or used.

In this PR, I implemented all the infrastructure necessary to write the metadata from chunked_array::ops and to read the metadata from polars-expr.

The writing is done using interior mutability. I attempted several approaches and this is the one that I settled on as being the easiest to work with and not providing to much overhead.

The reading is done through a trait object over the MetadataTrait, this makes all data available as Scalars.

The follow snippet illustrates that it works.

import polars as pl
import numpy as np
import os
from timeit import timeit

def many_mins():
    df = pl.DataFrame({ 'a': pl.Series(np.arange(500000)) })

    for i in range(10):
        df = df.with_columns(amin = pl.col.a.min())

os.environ["POLARS_METADATA_USE"] = "experimental"
with_md = timeit(many_mins, number=100)
os.environ["POLARS_METADATA_USE"] = "1"
without_md = timeit(many_mins, number=100)

print(f'without md = {without_md}')
print(f'with md    = {with_md}')

This outputs:

without md = 4.232960837998689
with md    = 0.6040126650004822