Datasets names follow up

[adonath]

Description

This pull request does some more clean up and adaption follow up work to #2782. The main changes included are:

• Remove Dataset.parameters, as all parameters are now in Dataset.models.parameters
• Remove the model section from the model setter and just select the right models in .npred() and .npred_sig()
• An important change that comes with this is that BackgroundModel now needs to define the datasets_names to be evaluated for the given dataset. We might re-consider this before v0.17, but for now it works like this.

[codecov]

Codecov Report

Merging #2785 into master will decrease coverage by 0.03%.
The diff coverage is 79.26%.

@@            Coverage Diff             @@
##           master    #2785      +/-   ##
==========================================
- Coverage   92.23%   92.19%   -0.04%     
==========================================
  Files         144      144              
  Lines       16548    16544       -4     
==========================================
- Hits        15263    15253      -10     
- Misses       1285     1291       +6

Impacted Files	Coverage Δ
gammapy/modeling/sampling.py	21.66% <0%> (ø)	⬆️
gammapy/time/lightcurve_estimator.py	96.1% <100%> (-0.1%)	⬇️
gammapy/cube/make.py	94.28% <100%> (ø)	⬆️
gammapy/cube/simulate.py	91.66% <100%> (ø)	⬆️
gammapy/modeling/datasets.py	86.33% <100%> (ø)	⬆️
gammapy/cube/fit.py	91.01% <100%> (-0.28%)	⬇️
gammapy/cube/fov.py	98% <100%> (ø)	⬆️
gammapy/spectrum/flux_point.py	93.19% <71.42%> (-0.26%)	⬇️
gammapy/spectrum/dataset.py	94.86% <72.22%> (-0.82%)	⬇️
gammapy/modeling/models/cube.py	91.69% <91.66%> (-0.06%)	⬇️
... and 2 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 bc9cc8c...7320fd5. Read the comment docs.