feat: add nested type constructor expressions

[jvanstraten]

It's currently impossible to formulate expressions like (field 0 - field 1, field 0 + field 1) (using (x, y) as a two-field struct):

• struct literals don't work because the fields aren't literals;
• mask expressions don't work because the fields aren't pure field references;
• extension functions don't work in general unless one is made specifically for structs with the desired number of fields, the desired nullability, and the desired type variation.

The same applies for lists and maps, though if we ignore type variations, the same could hypothetically be done for them with a finite number of function extensions in core.

This construct effectively makes struct/list/map literals and (I think) mask expressions redundant, but I didn't deprecate them because they at least feel different.

[jacques-n]

I'm generally supportive of this change (+1). It feels cleaner as an expression than trying to introduce as functions. (As functions, type declaration would get nasty quickly.)

@westonpace and @cpcloud , are you supportive of this?

[westonpace]

Took me a moment to figure this out but I think I get it now. I'm +1.

[CLAassistant]

All committers have signed the CLA.

[jacques-n]

Merging on lazy consensus