Add infinite-width bit extract operator

[albert-magyar]

Related issue: #857
Type of change: other enhancement
Impact: API addition (no impact on existing code)
Development Phase: hypothetical proposal

@jackkoenig @azidar @ducky64 This is an example of the hypothetical "infinite extract" operator. Not that I'm saying this is the right solution, but it is one easy solution.

Note: i is a terrible name that I just made up as an example.

[ducky64]

Can we do something more compositional rather than introducing a special-case extract operation? I'd prefer something like (5.U.pad)(4, 2), though we should think about that an unbounded pad would do (or whether it's even useful) in other cases.

[azidar]

I'm not sure what the final word on this was, but I personally am in favor of not breaking backwards compatibility for this, and just adding a safe extract operator. I don't like i as the API though....

[ducky64]

So the issue is that the backwards compatible solution means inconsistent behavior depending on whether your bit extraction is handled in Chisel frontend on a literal (in which case it implicitly pads the number to infinity) or in FIRRTL (where it would error out). And I'd argue that the former case is probably more of a bug than anything else, since it was a result of trying to special case literals in the chisel frontend but failing to maintain consistent behavior with FIRRTL.

We could also introduce a runtime deprecation phase, where the inconsistency will continue to work but emit a runtime deprecation warning, before removing it outright.