Add regression tests to ensure stable drop order

[aochagavia]

Work towards #43034

I think this is all we need to do on the testing front regarding RFC 1857

[rust-highfive]

r? @aturon

(rust_highfive has picked a reviewer for you, use r? to override)

[TimNN]

I think this is not entirely correct. As far as I know, here the fields are dropped in reverse order of initialization (not declaration) -- that is, If you swapped the lines with x: and y: below, x would be dropped before y. I think it would be good to have tests, for structs and enum-struct variants, which use different initialization and declaration orders.

[aochagavia]

You are right... I always confuse these terms... I am going to correct the comment and add more tests.

[aochagavia]

Just pushed two tests more to ensure different field initialization order behaves as expected

[arielb1]

I would consider merging this with the mir-drop-order.rs test, but it mostly tests drop glue/destructors rather than "MIR" drop order, so we can keep them separate. Maybe rename it to rfc1857-drop-order.rs so we can find it easily? r+ modulo that.

[aochagavia]

@arielb1 done!

[arielb1]

@bors r+ rollup

[bors]

📌 Commit f86e433 has been approved by arielb1