[stdlib] Introduce Hasher type with all necessary changes

[mzaks]

This is a draft because although the code compiles 8 tests are failing.
It might be due to compiler bug. The error messages are cryptic. I don't have the "Mojo" ;) to fix them.

Failed Tests (8):
Mojo Standard Library :: builtin/test_dtype.mojo
Mojo Standard Library :: builtin/test_hash.mojo
Mojo Standard Library :: builtin/test_object.mojo
Mojo Standard Library :: builtin/test_string.mojo
Mojo Standard Library :: builtin/test_string_literal.mojo
Mojo Standard Library :: collections/test_dict.mojo
Mojo Standard Library :: collections/test_set.mojo
Mojo Standard Library :: python/test_python_object.mojo

[mzaks]

Minimal example to reproduce the issue:
https://github.com/mzaks/mojo/tree/feature/minimal-example-of-test-crash-for-new-hasher

[mzaks]

@JoeLoser do we have a status on this one? From what I can tell there are some unexpected deeper language problems, my guess is it has something to do with register passable types. I am ok with shelving it for now. If so, I will concentrate on some other things and revisit when it is resolvable.

[mzaks]

Actually now I rebased the minimal reproducible example mzaks:feature/minimal-example-of-test-crash-for-new-hasher and it some times passes and some times fails the test. Seems like the order of tests is important and there is something fishy going on.

[mzaks]

After rebasing the mzaks:feature/minimal-example-of-test-crash-for-new-hasher on todays nightly, I see following compile error on :

invalid indirect call: l-value of type '$0' cannot be converted to reference of type 'H'
struct Int(
       ^~~

The H in question is from:

trait NewHashable:
    fn __hash_with__[H: Hasher](self, inout hasher: H):
        pass

which Int is implementing.

I pushed the rebased version so the team can confirm.