Fix compute rms

[matteobachetti]

• Make the code for rms computation a little more clear
• deprecate old behavior
• Check documentation --> Small update to Spectral Timing Exploration notebook

I also had to fix a couple of things related to warnings, because the new pytest is much pickier. All the changes to warnings (and to imports in the top of files) are due to that. Please ignore for the sake of this PR. Only consider changes to (test_)crossspectrum, powerspectrum, fourier, and minor related changes in multitaper, lombscargle and deadtime.

Resolve #790

[codecov]

Codecov Report

Attention: 1 lines in your changes are missing coverage. Please review.

Comparison is base (6f38ba2) 78.98% compared to head (02fe583) 95.83%.

❗ Current head 02fe583 differs from pull request most recent head 10764e4. Consider uploading reports for the commit 10764e4 to get more accurate results

Files	Patch %	Lines
stingray/crossspectrum.py	87.50%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##             main     #792       +/-   ##
===========================================
+ Coverage   78.98%   95.83%   +16.84%     
===========================================
  Files          43       43               
  Lines        8762     8784       +22     
===========================================
+ Hits         6921     8418     +1497     
+ Misses       1841      366     -1475     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[pep8speaks]

Hello @matteobachetti! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

• In the file stingray/tests/test_lightcurve.py:

Line 1713:101: E501 line too long (101 > 100 characters)
Line 1734:101: E501 line too long (101 > 100 characters)

Comment last updated at 2024-01-29 14:44:59 UTC

[mgullik]

Hi @matteobachetti,

very nice! This will make the rms calculation much stronger, especially the error estimate.

I noticed that in the tests of get_rms_from... you build your synthetic PDS from a model shape and then randomising the values. Is this procedure different from simulating the light curve from the same theoretical shape of the PDS (with Timmer&Koenig method) and calculating the PDS from the light curve?
Does it make sense to perform all the tests using this last way? It seems closer to the usual procedure that calculates the rms.