Instead of -max_size to -1 for regular refcounts, use 1 to max_size.
0 still means constant refcount.
The highest bit is used to signify atomic refcounting required.
This does not turn on any sort of atomic refcounting.
MulSaturated and AddSaturated are now implemented for u128, i128
SubSaturated is now implemented for Dec even if it's a little strange to
saturate at a decimal value:
```
» Num.subSaturated -170_141_183_460_469_231_731dec 1
-170141183460469231731.687303715884105728 : Dec
```
I decided to rm the `build_num_{mul,add}_saturated` methods because
they don't require any asm-specifics. They either call out to bitcode, or
use the non-saturated version for floats.
Dec negation was implemented across gen-dev, gen-llvm, gen-wasm as a call
to the compiled zig function `bitcode::DEC_NEGATE`.
f32 and f64 negation were implemented already for gen-llvm, gen-wasm.
for gen-dev x86_64, float negation is implemented by flipping the sign
bit, which means `xorps` for f32, and `xorpd` for f64
for gen-dev aarch64, there is conveniently a `fneg` instruction
This was never fully hooked up in platforms, and the plan is to replace the need for this with doing purity-inference on normal `expect` statements.
On the other hand, fuzzing is finding some bugs caused by having a hyphenated keyword, so this is a great time to go ahead and remove it!
Currently, list literals are always heap-allocated and their elements
are stored by emitting a GEP and store for each item in the literal.
This produces huge quantities of IR, causing compile times for e.g.
programs with large literals or ingested files to blow up.
Instead, if a list literal consists entirely of literal values, create a
global section for the literal and return a pointer to it.
