keep order of provided object constructor fields in sem

[metagn]

closes #1056 (for the sem issue, evaluation order is still not fixed)

The compiler reorders the fields of an object constructor depending on the field order of the object type, this PR makes it so that provided fields are kept in the order that they were provided and only the default fields keep the object type order. This still encounters UB in the generated C so it doesn't fix the evaluation order, but it might still be useful, and things like:

Foo(
  b: (let x = 123; x),
  a: x)

won't break on resem.

[metagn]

Ended up not fully fixing the issue, could close and postpone this for compat mode.

[Araq]

I think sem should produce a warning or an error if there are function calls involved and the order of the fields differs from the declaration order. I don't see the benefit in allowing an arbitrary order in 2025 when auto-completion is now so good it threatens our jobs anyway.