Tweak in MapEvaluator.need_update for performance

[QRemy]

This PR introduce a performance tweak in MapEvaluator.need_update that replaces skycoord.separation by angular_separation in order to avoid unnecessary frame and unit equivalence comparison.

[codecov]

Codecov Report

Merging #3353 (af78a32) into master (4e1c72b) will decrease coverage by 0.02%.
The diff coverage is 88.23%.

@@            Coverage Diff             @@
##           master    #3353      +/-   ##
==========================================
- Coverage   93.72%   93.70%   -0.03%     
==========================================
  Files         144      144              
  Lines       18270    18282      +12     
==========================================
+ Hits        17124    17131       +7     
- Misses       1146     1151       +5     

Impacted Files	Coverage Δ
gammapy/datasets/map.py	94.34% <88.23%> (-0.13%)	⬇️
gammapy/modeling/iminuit.py	96.73% <0.00%> (-3.27%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4e1c72b...af78a32. Read the comment docs.

[QRemy]

Added a test to avoid update if there is no free spatial parameters (solve fail related to TemplateSpatialModel)

[adonath]

Thanks a lot @QRemy! One last suggestion to try: maybe it is an option to also check for if parameters_spatial_changed in .need_update and only check for the separation if this is the case?

[adonath]

Thanks @QRemy, it seems there are now test fails https://github.com/gammapy/gammapy/pull/3353/checks?check_run_id=2557044541#step:9:1865, I can also reproduce them locally....not sure what is happening. Maybe see if they can be fixed with limited effort...

[adonath]

Thanks @QRemy, the remaining test fails seems unrelated. I think adding the lon_0 and lat_0 parameters to the TemplateSpatialModel is still a good idea, however this can be done in a follow up PR.