Add a fallback on annotation validation

[kbolashev]

Having a malformed annotation on a datapoint (or an annotation that failed to parse, not even malformed) right now leads to the whole loading process failing.

This PR is a hotfix to prevent that.
If an annotation fails to parse, it gets shown to the user in a warning message, and I warn them that they won't be able to interact with them, for example:

Accessing as kirill
Downloading annotation fields ['annotation']...
Warning: The following datapoints had invalid annotations, any annotation-related operations will not work on these:
Field "annotation" in datapoints:
        train2017/000000000009.jpg

I'm not handling these bad annotations in other downstream functions for now properly, if we do want to make it have a nice UX, would probably need to have some kind of BrokenAnnotation class that will shim the regular annotation functions and display a warning.
Not sure if it's worth it to work on this right now

[dagshub]

Join the discussion on DagsHub!

[EvgeniLeonti]

LGTM

Make sure formatting is ok:

dagshub/data_engine/model/query_result.py#L446
Too many blank lines (2) (E303)

[kbolashev]

Good catch, I think my autoformatter broke 🥲