review the possibility to add invalid hits in track candidates passed by mkFit to the final fit

[slava77]

Currently mkFitOutputConverter drops all non-valid hits in the produced TrackCandidates. Some of the useful information is lost: we could pass invalid,bad, or other status hits that could be potentially useful in the final fit.

Most of this will likely require additional level of detail added inside of mkFit, e.g.,

• ability to provide a DetId per dead hit
• ability to provide a full hit index for potentially useful hits which were skipped in the pattern reco
  • currently we just assign a -x to the hit index; -> hit index and hit status would need to be tracked separately

[slava77]

I have filled more details for this issue.
(nb, this is a follow up to the Apr 20 UCSD meeting)

[slava77]

since this issue was opened, a couple of points to motivate to add the invalid hits to the output candidates (TrackCandidate)

• without an invalid hit in the track pattern the final fit does not re-discover a missing hit and would not use the missing hit/layer to update the material properties
• lack of lost hits in the trajectory does not allow to reject tracks during track selection/classification
• there is a tracker hit efficiency measurement method based on checking for (in)valid hits in the track trajectory (between existing hits); this method does not work anymore out of the box