Distinguish the cases of "zero files in dataset", "number of files is unknown", and "dataset does not exist"

[ponyisi]

There seem to be multiple places in the user interface (e.g. the dashboard) where a dataset having zero files (e.g. when a Rucio dataset does not exist) is reported as "Unknown".

In the case of Rucio DIDs this is often because the dataset does not exist, which we should handle in some more obvious way than reporting an "Unknown" number of files. In some cases the Rucio dataset might exist but be empty, in which case we should properly report zero files.

[ponyisi]

Also add the case "DID finder crashed"...

[ponyisi]

Proposal:

Database models:

• the DatasetStatus enum gains the extra value lookup_failed
• the TransformStatus enum gains the extra value bad_dataset; this will be considered to be a termination status for the transformation. (It will require client support ...)

We add another status field to the JSON sent to FilesetComplete. If the lookup is a success the field will say so. If the DID finder decides that the input DID specification is invalid, this field will mark that, and if the DID finder crashed in the lookup, there will be a third marker value.

Based on this:

• if the DID finder decides the DID is bad it will raise an exception (specified in the DID finder library) which the DID finder library will catch and use to return the "invalid DID" status. The ServiceX app will then mark the dataset with lookup_failed, and all referring transforms with bad_dataset. The dataset will be marked as stale (so in case it does exist in the future we will be able to find it).
• if the DID finder fails otherwise the library will return the "lookup failure" status, at which point the dataset will be marked as lookup_failed and stale, and referring transforms will be marked as "fatal" completion.

[gordonwatts]

See #1134 for the infrastructure - partially completes this.