Currently we uassert if the backup cursor metadata has no 'checkpointTimestamp' field, which is unlikely but could happen if the sync source hasn't taken a stable checkpoint yet. Instead of uasserting we should fail the FCBIS attempt (so we can retry FCBIS, since likely the sync source will have taken a stable checkpoint later on)

Additionally we should audit tests that use FCBIS and consider using `awaitLastRecoveryTimestamp` to make sure the sync source has taken a stable checkpoint before we start FCBIS.

We should also look into whether other usages of _getBSONField in FCBIS should retry instead of uasserting.