The compiler checks for exhaustive match arms in match expressions.

[emilyaherbert]

Closes #581

This PR does two things:

1. I had to introduce a pretty significant change to the way Expressions are parsed. In order to reliably gather type information for exhaustivity checking, all "primary expressions" inside of match expressions must first be named as variables, then that new generated variable name is used within the match expression. The problem is that before this PR, match expressions could not declare variables as they had no access to creating AstNodes. This PR introduces a new type ParseResult that "bubbles up" a Vec that is populated by match expressions. Moreover, this change fixes a silent bug where if the value of your match expression was a function with side effects, that function would have been called n number of times for n match arms.
2. Implement match exhaustivity checking and match arm reachability checking. There is a bit of math involved, but I describe it at a high level in sway-core/src/semantic_analysis/ast_node/expression/usefulness/mod.rs. Let me know if you have questions 😄

[emilyaherbert]

Pinging @canndrew @sezna @mohammadfawaz @otrho for review 😄

[sezna]

Just one question, then I'm ready to approve: #701 (comment)

[adlerjohn]

Looks like @otrho won the merge conflict race

[emilyaherbert]

Pinging @canndrew @sezna @mohammadfawaz @otrho for review 😄