cache=True (default) in to_datetime causes too big slow-down

[MarcoGorelli]

Checks

• I have checked that this issue has not already been reported.
• I have confirmed this bug exists on the latest version of Polars.

Reproducible example

s = pl.datetime_range(date(2000, 1, 1), date(2001, 1, 1), '1s', eager=True)
strings = s.dt.strftime('%Y-%m-%dT%H:%M:%S')

# The following takes > 10 seconds
strings.str.to_datetime(format='%Y-%m-%dT%H:%M:%S')

# The following takes < 1 second
strings.str.to_datetime(format='%Y-%m-%dT%H:%M:%S', cache=False)

Log output

No response

Issue description

I don't think the cache should hurt performance - I get the desire to speed it up in the case of duplicates, but unique datetimes is a pretty common case too

Is it possible to limit the cache so that it only holds, say, 50 elements? And then, any new element ejects an existing one from the cache?

Expected behavior

Similar performance

Installed versions

In [12]: pl.show_versions()
--------Version info---------
Polars:               0.20.21
Index type:           UInt32
Platform:             Linux-5.15.146.1-microsoft-standard-WSL2-x86_64-with-glibc2.35
Python:               3.11.9 (main, Apr  6 2024, 17:59:24) [GCC 11.4.0]

----Optional dependencies----
adbc_driver_manager:  <not installed>
cloudpickle:          3.0.0
connectorx:           <not installed>
deltalake:            <not installed>
fastexcel:            <not installed>
fsspec:               2024.3.1
gevent:               <not installed>
hvplot:               0.9.2
matplotlib:           3.8.4
nest_asyncio:         1.6.0
numpy:                1.26.4
openpyxl:             <not installed>
pandas:               2.2.2
pyarrow:              15.0.2
pydantic:             <not installed>
pyiceberg:            <not installed>
pyxlsb:               <not installed>
sqlalchemy:           <not installed>
xlsx2csv:             <not installed>
xlsxwriter:           <not installed>

[ritchie46]

I think FastFixedCache should be used in CachedFunc. Currently it is an unlimited hashmap.

[reswqa]

As we discussed, we should use FastFixedCache to replace the hashmap in CachedFunc, and the size can be sqrt(ca.len()). If no one beats me, I'll file a PR then.