Follow up for SERVER-87215 which added more logging

On next occurrence we should have more info to debug

Theories considered as root cause:

• BSON lifetime issues
  • Possible, but no obvious misuse
• Threading issues
  • There are multiple validation runs going on at once, but there should be no overlap in the data inspected
• Conversion from RecordData -> BSON
  • If a null is appended anywhere it could explain this
  • Would be more widespread if this were an issue
• Some storage engine issue where the returned value is broken in some way
  • This would be super weird, especially since we seek the same record id at an earlier time with no issue

Other notes:

• Issue does not trivially reproduce, but seems to be common enough that there is clearly some environmental condition leading to it, we have a long history of samples exhibiting this behavior across major versions. Likely also related to the key being missing in the first place?
  • This issue is not yet confirmed in 8.1, after a large number of validation refactors. It's unclear if the issue is actually resolved due to a lack of sampled (most failed 8.1 validations don't have missingIndexEntries populated

• Using validation run 682728d76ff1d05bf4ccab8a as an example from here, recordId 38376 is listed as missing, but when printing the metadata, there is a record with that recordId and a valid _id field.
• Seen on a few text indexes, but also on geo
• The _id field is clearly present in the returned data obj, otherwise it could not be added
• The recordIds being read are in the collection, and printing the metadata shows intact _id fields