Add a second cpp2::range constructor which takes 2 types, and deduce the common type for them (if available)

[bluetarpmedia]

Use a CTAD deduction guide to deduce the std::common_type for both types (if a common one exists).

This allows code like the following to compile without integer precision conversion warnings:

for 0 ..< vector.size() do (e) { ... }

since the integer literal 0 and the size expression evaluate as different signed/unsigned types.

Also update the regression tests after recent changes.

Fixes #1269

[feature-engineer]

Does putting e - 1 < 0 inside the loop still fails to compile (as it should)?
What about vector[e - 1] when e is zero?
Or when we iterate backwards?

[hsutter]

Thanks!

@feature-engineer AFAICS this only affects ..< and ..= expressions so those would be unchanged.