Further improve model region / mask selection

[adonath]

Description
This pull request is a follow-up to #3169. It further breaks down the functionality like so:

• Introduce Map.mask_nearest_position(), which returns the nearest position inside the mask to given position
• Move the selection of the nearest IRF to IRFMap. Right now it was using the mask_safe_psf for PSF as EDisp, however it's not guaranteed that those agree. For this is introduce a IRFMap.mask_safe_irf, which derives the mask from the data. however we might think about moving the mask safe for the IRF to the IRFMap completely, this would be more than convenient for stacking as well...
• Test wise I introduced RegionGeom.from_regions(), which creates a RegionGeom with a CompoundSkyRegion from a list of regions, while working on this I think we should clarify the handling of CompoundRegion vs. a list of Region, everywhere in Gammapy. I added it as a discussion point on the agenda for the dev call on Friday.

Dear reviewer

[codecov]

Codecov Report

Merging #3215 (fb3e2ba) into master (39133de) will decrease coverage by 0.03%.
The diff coverage is 91.83%.

@@            Coverage Diff             @@
##           master    #3215      +/-   ##
==========================================
- Coverage   93.91%   93.87%   -0.04%     
==========================================
  Files         144      144              
  Lines       17441    17467      +26     
==========================================
+ Hits        16380    16398      +18     
- Misses       1061     1069       +8     

Impacted Files	Coverage Δ
gammapy/datasets/simulate.py	99.40% <ø> (ø)
gammapy/estimators/ts_map.py	97.66% <ø> (ø)
gammapy/maps/wcsnd.py	95.31% <75.00%> (-0.79%)	⬇️
gammapy/modeling/models/cube.py	87.98% <81.25%> (-0.51%)	⬇️
gammapy/maps/core.py	86.56% <87.50%> (+0.01%)	⬆️
gammapy/maps/region.py	89.61% <88.88%> (-0.07%)	⬇️
gammapy/datasets/map.py	94.41% <100.00%> (-0.14%)	⬇️
gammapy/irf/core.py	91.10% <100.00%> (+0.35%)	⬆️
gammapy/irf/edisp_map.py	96.05% <100.00%> (+0.05%)	⬆️
gammapy/irf/psf/psf_map.py	96.62% <100.00%> (+0.02%)	⬆️
... and 3 more

---

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 39133de...fb3e2ba. Read the comment docs.