Full likelihoods

[jackaraz]

Description of the Change:
add the ability to include CRs and VRs in the full likelihoods.

Benefits:

• In general, ATLAS analyses are designed such that the signal contamination of CRs is max 10%. So recasting only the SRs but not the CRs/VRs is a good approximation in most cases (as long as the signal hypothesis is not very different from the one the analysis has been designed for).
• Some analyses do a combined fit to signal and control regions; others don't. If they do, removing the CRs altogether is wrong. If they don't, it makes no difference whether or not the CRs are kept.

New options:
Option introduced in the .info file of the analysis e.g.

  <pyhf id="RegionC">
    <name>atlas_susy_2018_31_SRC.json</name>
    <regions>
      <channel name="CRtt_cuts"> </channel>
      <channel name="SR_metsigST">
      SRC_22
      SRC_24
      SRC_26
      SRC_28
      </channel>
      <channel name="VRzAlt_cuts" is_included="True"> </channel>
      <channel name="CRz_cuts" is_included="True"> </channel>
      <channel name="VRtt_cuts"> </channel>
      <channel name="VRz_cuts"> </channel>
    </regions>
  </pyhf>

If there is no data present in the channel code will look for the is_included tag, which can be "True", "yes", "1". If true, the channel will be included in the likelihood calculation; if not, it will be removed. If is_included is not present, and the channel does not have any data, it will be removed by default.