Improve schema evolution testing

[makortel]

Motivated by cms-sw/cmssw#41246

For each class that is part of RAW data tier, craft a test where we read in a ROOT file (stored in https://github.com/cms-data/) with one Event that has an object that is "fully filled" (all member containers have at least one element). Add a README.md for each DataFormat package holding such class explaining the test, and a policy that whenever such class is updated, a new test file needs to be added whenever the class or any of its contained classes are changed. Maybe also craft a test where one cmsRun writes a file with the object and another reads file, both in the same release. Currently RAW data tier includes

• Add tests to be able to read back old version(s) of FEDRawDataCollection #541
• Add tests to be able to read back old version(s) of GlobalObjectMapRecord #542
• Add tests to be able to read back old version(s) of edm::TriggerResults #540
• Add tests to be able to read back old version(s) of trigger::TriggerEvent #543

Add similar tests for Scouting data formats

• Make one file for all Run 2 scouting classes, and another file for all Run 3 scouting classes. Consider encoding the class versions of all of the participating classes in the file name. Policy: whenever any of RunX scouting class is updated, create a new file, and if the data format used in the most recent file has not been used in production, delete that file.
• This development would also resolve Add tests or workflows that read existing scouting data cmssw#41040 and [ROOT] Unit Tests testDataFormatsScoutingRunX failing in ROOT IBs cmssw#41222

• For Run3 Scouting classes do similar test as for RAW #544
• For Run2 Scouting classes do similar test as for RAW #572
• Clean up previous Scouting tests #576

Add similar tests for RAW' data formats

• For SiStripApproximateClusterCollection add similar tests as for RAW #676
  • After Add SiStripApproximateClusterCollection as a simple format for RAW (re-vamped) cmssw#42495 gets merged
• For DetId and std::vector add similar tests as for RAW #677
  • This is pending for what happens after Add SiStripApproximateClusterCollection as a simple format for RAW (re-vamped) cmssw#42495 (comment)
  • After use std::vector<DetId> as data format of SiStrip DetIds with FED errors cmssw#42662 gets merged

Finally

• Craft our own schema evolution test suite. #545
  • Try to exercise as much of the structure of nested containers as feasible to mimic the complexity of our data formats. Tricky part is to build the "schema evolved" data formats. https://github.com/cms-sw/cmssw/blob/master/IOPool/Input/test/testNoParentDictionary.sh may be helpful

Not really about schema evolution testing, but to track in this issue the earlier milestone on "responding to schema evolution failure"

• Develop workaround to read files not having all necessary StreamerInfo #534

Progress tracked in https://github.com/makortel/framework/milestone/31