Add TCAppL and TCAppR typechecking stack frames, and remove TCBind{L,R}

[byorgey]

Towards #1314.

• It seems like it could be helpful to report that we were checking the left- or right-hand side of a function application when reporting type errors, to help give context.
• On the other hand the "checking the LHS/RHS of a semicolon" never seemed very useful, since it would always come in a big chain and didn't really help localize the error, so I removed it.
• I'm open to suggestions for other context we could provide. There is definitely a balance here between providing helpful context and providing too much. For example, should we add something that says "while checking the LHS/RHS of a pair"?

[xsebek]

Nice, the error message for a function argument looks better. 👍

What happens with id 3 3? Is that a RHS of LHS? 🤔

[byorgey]

What happens with id 3 3? Is that a RHS of LHS? 🤔

id 3 3 reports that id 3 was expected to have a function type but actually has type Int, while checking the LHS of the application _ 3.

Maybe an example of what you were thinking of would be something like "hi" + 2, which now reports

1:1: Type mismatch:
  From context, expected `"hi"` to have type `Int`,
  but it actually has type `Text`

  - While checking the argument to a function: (+) _
  - While checking a function applied to an argument: _ 2