Use inverse transform when comparing scores in fpca tests

[Ddelval]

Compare the functions obtained with the inverse_transform instead of the scores obtained in transform directly. In doing so, we can use a smaller tolerance (0.25 before vs 0.1 now).

There are two things to keep in mind:

• The internal weights parameter of FPCA is changed between the calls to fit and transform because fda.usc uses different quadratures for these two operations.
• Using inverse_transform with coefficients obtained from fda.usc is not strictly correct since the principal components calculated in fda.usc are slightly different. However, the result is very similar, and this difference is not the cause of still having to use a high tolerance (0.1).

[codecov[bot]]

Codecov Report

Base: 85.28% // Head: 85.28% // No change to project coverage :thumbsup:

Coverage data is based on head (e22beeb) compared to base (e22beeb). Patch has no changes to coverable lines.

:exclamation: Current head e22beeb differs from pull request most recent head 168dac6. Consider uploading reports for the commit 168dac6 to get more accurate results

Additional details and impacted files

```diff @@ Coverage Diff @@ ## develop #484 +/- ## ======================================== Coverage 85.28% 85.28% ======================================== Files 137 137 Lines 11040 11040 ======================================== Hits 9415 9415 Misses 1625 1625 ``` Help us with your feedback. Take ten seconds to tell us [how you rate us](https://about.codecov.io/nps?utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=GAA-UAM). Have a feature suggestion? [Share it here.](https://app.codecov.io/gh/feedback/?utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=GAA-UAM)

:umbrella: View full report at Codecov.
:loudspeaker: Do you have feedback about the report comment? Let us know in this issue.