feat(python): improve Series/DataFrame init from existing Series/DataFrame objects

[alexander-beedie]

Closes #13338.

Various init enhancements. Can already init Series and DataFrame objects directly from an existing Series; now we can additionally init both DataFrame and Series from an existing DataFrame (improving cross-class constructor consistency).

Also now preserves Series name on Series(Series) init, and adds some (marginal) fast-paths if neither "schema" nor "schema_overrides" params are passed to the DataFrame constructor when initialising from either type (if these parameters are passed, they are handled appropriately).

Examples

import polars as pl

df1 = pl.DataFrame( {"id": [0, 1], "misc": ["a", "b"], "val": [-10, 10]} )
# ┌─────┬──────┬─────┐
# │ id  ┆ misc ┆ val │
# │ --- ┆ ---  ┆ --- │
# │ i64 ┆ str  ┆ i64 │
# ╞═════╪══════╪═════╡
# │ 0   ┆ a    ┆ -10 │
# │ 1   ┆ b    ┆ 10  │
# └─────┴──────┴─────┘

df2 = pl.DataFrame( df1, schema=["a", "b", "c"], schema_overrides={"val": pl.Int8} )
# ┌─────┬─────┬─────┐
# │ a   ┆ b   ┆ c   │
# │ --- ┆ --- ┆ --- │
# │ i64 ┆ str ┆ i8  │
# ╞═════╪═════╪═════╡
# │ 0   ┆ a   ┆ -10 │
# │ 1   ┆ b   ┆ 10  │
# └─────┴─────┴─────┘

s1 = pl.Series( "s", df2 )
# Series: 's' [struct[3]]
# [
#   {0,"a",-10}
#   {1,"b",10}
# ]

s2 = pl.Series( s1 )
# Series: 's' [struct[3]]
# [
#   {0,"a",-10}
#   {1,"b",10}
# ]

[stinodego]

I like the new functionality! Just one minor nitpick and this can be merged 👍