Fix temp local confusion in field assignment

[dcodeIO]

This PR fixes an issue caused by indexes of potential temporary locals in the RHS of a field assignment not being blocked preemptively. Also reorders the arguments to makeReplace to adhere to execution order fwiw.

fixes #1095

[MaxGraey]

Could you plz add also minimal reproduction example for prevent regression in future?

[dcodeIO]

Yeah, was once again wondering what to do with this one. Is relatively specific to a managed temp. in the RHS, and making the assumption that a particular pattern uses a temp for something managed feels wonky, since unrelated changes to how the RHS is compiled might break it easily.

What do you think of creating a test directory for regression tests specifically, so we can have tests like issue/1109.ts?

[MaxGraey]

Sounds great! It specially important when we have major update/refactoring of compiler