feat(python, rust): clearer message when stringcache-related errors occur

[MarcoGorelli]

related to #9106 (debatable whether it closes it)

Demo:

latest release

In [2]: df1 = pl.DataFrame({"a": ["1", "2"]}, schema={"a": pl.Categorical})
   ...: df2 = pl.DataFrame({"a": ["1", "2"]}, schema={"a": pl.Categorical})
   ...: df1.join(df2, on='a')
---------------------------------------------------------------------------
ComputeError: joins/or comparisons on categoricals can only happen if they were created under the same global string cache

In [3]: df1 = pl.DataFrame({"a": ["1", "2"]}, schema={"a": pl.Categorical})
   ...: df2 = pl.DataFrame({"a": ["3", "4"]}, schema={"a": pl.Categorical})
   ...: pl.concat([df1, df2])
---------------------------------------------------------------------------
ComputeError: cannot concat categoricals coming from a different source; consider setting a global StringCache

here

In [1]: df1 = pl.DataFrame({"a": ["1", "2"]}, schema={"a": pl.Categorical})
   ...: df2 = pl.DataFrame({"a": ["1", "2"]}, schema={"a": pl.Categorical})
   ...: df1.join(df2, on='a')
---------------------------------------------------------------------------
ComputeError: joins/or comparisons on categoricals can only happen
if they were created under the same global string cache.

Help: if you're using Python, this may look something like:

    with pl.StringCache():
        df1 = pl.DataFrame({'a': ['1', '2']}, schema={'a': pl.Categorical})
        df2 = pl.DataFrame({'a': ['1', '3']}, schema={'a': pl.Categorical})
        df1.join(df2, on='a')

Alternatively, if the performance cost is acceptable, you could just set:

    pl.enable_string_cache(True)

on startup.

In [2]: df1 = pl.DataFrame({"a": ["1", "2"]}, schema={"a": pl.Categorical})
   ...: df2 = pl.DataFrame({"a": ["3", "4"]}, schema={"a": pl.Categorical})
   ...: pl.concat([df1, df2])
---------------------------------------------------------------------------
ComputeError: cannot concat categoricals coming from a different source, consider setting a global StringCache.

Help: if you're using Python, this may look something like:

    with pl.StringCache():
        df1 = pl.DataFrame({'a': ['1', '2']}, schema={'a': pl.Categorical})
        df2 = pl.DataFrame({'a': ['1', '3']}, schema={'a': pl.Categorical})
        pl.concat([df1, df2])

Alternatively, if the performance cost is acceptable, you could just set:

    pl.enable_string_cache(True)

on startup.

[ritchie46]

Nice! Thanks a lot!

One remark. I think we can store this message in polars-error under a shortcut (there are more).

[ritchie46]

Can we deduplicate this message? Maybe add it in polars-error?