Improve hashing of function and pseudocode objects

[d-torrance]

We add hash codes to all members of the Code union in the interpreter, which we then use to compute proper hashes of function and pseudocode objects.

Before

i1 : hash (() -> print "Hello, world!")

o1 = 6474929089309016109

i2 : hash (() -> print "Hello, world!")

o2 = 14328623249302651656

After

i1 : hash(() -> print "Hello, world!")

o1 = 9105046539553097294

i2 : hash(() -> print "Hello, world!")

o2 = 9105046539553097294

Closes: #667

AI Disclosure

This was almost entirely vibe-coded by Claude Code, but with serious hand-holding.

[pzinn]

Are you saying Claude Code can program in "d", despite the lack of any meaningful documentation?

[d-torrance]

Yup! It was basically just copying and modifying the hashing function for the Expr union, but it did a decent job!

[pzinn]

Interesting. I for one would be happy to never have to write another line of "d".
BTW:
Maybe that's something you guys can talk about at the upcoming M2 meeting: how to get Claude Code to entirely rewrite the whole d code into a more modern language...

[mahrud]

Would it be feasible to have a single source of truth for all magic numbers?