for 2.0, make try/catch/else/finally scopeless?

[StefanKarpinski]

There's been a number of complaints over the years about the fact that each clause of a try/catch/else/finally construct introduces new local scope. This means that if you assign something in the try clause it won't be visible afterwards. But moreover, it won't be visible in the catch, else or finally blocks. The lack of visibility in catch block is and after the whole try construct is somewhat justifiable since we don't want to by default leave undefined variables (although it is easy enough to do with something like if cond(); x = 1; end. Perhaps it would be simpler and easier if the whole try construct didn't introduce any local scopes? Similar to how if/else doesn't. Just a thought. Reference example issue: #46928.

[IanButterworth]

Aside from changing the default behavior, triage approved of a @noscope macro (#39217) for making a try etc. work like this, but I stalled because I think to do it properly requires deeper understanding of the interpreter

[PallHaraldsson]

try-clauses are pretty fundamental, and something we want the right semantics for. While this is breaking, would a lot of code out there likely NOT break (such as none in the standard library)? [And with this change would a macro for the opposite, the old, behavior be needed, except for compatibility?]

Do we want to do Julia 2.0, sooner rather than later because of this? I guess for a while we would want a switch to opt into older behavior (for the same Julia 2.x binaries and/or display a warning for some code where behavior will change).