Introduce PvaAbstractions and use them in SeqTable

[evalott100]

Closes #442 and #310

[coretl]

As discussed, please could we have:

SignalType = TypeVar("SignalType", bool, int, ..., BaseModel)

class Signal:
    def __init__(self, datatype: Optional[SignalType] = None):
        check_one_of(datatype, SignalType)
        ...

I forgot that we wanted to take datatype in Signal.__init__ so that we could create a Signal with a type but no backend in create_children_from_annotations, then fill it in on connect.

As per #310 (comment)

[evalott100]

@coretl

Can we implement #310 (comment) in another issue/PR? This is already a very big change, and changing init/connect args for Signal/SignalBackend touches so many places in the code-base.

Additionally, I'd like to disallow arg -> kwarg inference on Signal and SignalBackend init and connect which we could do in the same PR.

[coretl]

@coretl

Can we implement #310 (comment) in another issue/PR? This is already a very big change, and changing init/connect args for Signal/SignalBackend touches so many places in the code-base.

Additionally, I'd like to disallow arg -> kwarg inference on Signal and SignalBackend init and connect which we could do in the same PR.

Sounds like a good idea, let me know when this needs another review

[evalott100]

New issue here. I'm ready for a new review #562

[coretl]

... as we have an implicitly defined repeats field that will have changes defaults from 1 to 0. Are we lacking a test for this plan stub?

[evalott100]

We do have tests for this but at no point in them do we check on repeats