Finish case related warnings

[MikePopoloski]

• Check completeness of small-sized bit vectors?
• Think about how priority, unique, unique0 apply
• Check applicability to all other forms of case statements
  • property exprs
  • pattern cases
  • randseq cases
  • case generates

[spomatasmd]

Recycling this issue as it seems related to what I'm experiencing: when using -Wcase-default, the message

warning: 'case' missing 'default' label

is correctly not produced for unique cases, but it is instead for the unique0 modifier. There should really be little to no difference at parsing time between the two (apart from obvious different simulation behaviour where no matches are allowed for unique0).

This happens with the latest slang-server v0.2.0 which uses slang as of 09333cb.

[MikePopoloski]

This is intentional. unique cases report an error in simulation if there's no match, so there's no reason to specify a default. unique0 does not, so you may still wish to have a default case to handle that. This is just an optional warning, you don't have to enable it.

[spomatasmd]

Fair, indeed it is an optional warning however the slang-server LSP enables it and that's how I stumbled onto it.
Still not 100% clear why would one pick unique0 with a default, I was under the assumption that, from an inferred hardware point of view, unique and unique0 were equivalent.

Need to dust off the document by C. Cummings since it's been a while from the last read and this periodically pops up every few months or so...