fix(save_df): include keyword names in the cache key

[LemonCondor]

Fixes #310

save_df built the cache file name from keyword argument values only, joined without a separator, so different calls could share a cache file — e.g. get_series(id, endPeriod="2023-01") returned the data cached by an earlier get_series(id, startPeriod="2023-01").

Change

• The key now uses name=value for keyword arguments, sorted by name (so keyword order doesn't create duplicate files), and parts are joined with a separator (\x1f) so positional boundaries can't collide (("AB", "C") vs ("A", "BC")).
• Existing cache files get a new name, so they are simply re-downloaded once.

Tests

tests/utils/test_save_df.py (offline, uses a temporary cache folder):

• startPeriod-like vs endPeriod-like keyword with the same value → distinct results (failed before)
• positional arguments with the same concatenation → distinct results (failed before)
• identical calls with keywords in a different order → one cache file

Also checked against the live API: after the fix, the example in #310 returns 397 rows (1990-01 .. 2023-01) as expected. tests/utils passes locally; black and flake8 pass with the project config.

[tfardet]

Thanks for the fix, I'm trying to see whether @hadrilec can setup an environment so we can run the tests, will try to come back to you soon

[tfardet]

I ran the test locally and everything looks good, thanks!