Add infrastructure to record skipped mutations for an index

There can be cases where we skip processing a certain mutation for an index. Eg: due to large array explosions, value is a scalar field but a vector field was expected, etc

It is better to record such skipped mutations/document IDs for long term and report them to user as and when required. Note: raising an alert for each skip maybe not be ideal as if we have some 1000s of mutations being skipped then user will get bombarded with alerts.