Remove Asrt.Star (fix #159)

[N1ark]

Remove Asrt.Star; what used to be Asrt.t is now Asrt.atom (atomic assertions), and Asrt.t is now atom list, avoiding the need for recursive functions when traversing assertions, and making the distinction much clearer and simpler.
Rename Asrt.GA -> Asrt.CorePred

[giltho]

I think if you fix this problem, it will also fix #301
This PR also re-orders the fields and it breaks something

[N1ark]

I think if you fix this problem, it will also fix #301 This PR also re-orders the fields and it breaks something

Ah yes interesting you seem to have stumbled upon similar things... When I have more time (🫠) I'll look; what's odd is that the version that works doesn't sort the assertions, so I'm not certain how sorting then (Types>Pure>CorePred>Wand) causes the issue; will update

And yes sure I'll rename it to atom it's nicer

Additional note: running Gillian-JS with this causes a stack overflow so something somewhere isn't tail-recursive anymore...

[N1ark]

JS works but something is still odd for Gillian-C, which ends up verifying less branches than it used to, possibly due to ordering issues. Will leave as is for now and maybe come back later to it...

[N1ark]

The tests pass which is a miracle but from diffing some of my local tests I get a few mismatches so not ready to merge yet fyi

[N1ark]

@giltho @NatKarmios fyi this is ready for review :)

After some evaluation, I get:

• 4% perf improvement on Collections-C
• AWS passes
• One specific proc is ~3 times slower, because more branches (77 rather than 13) are covered than before. I checked and they're ok, it's caused by some assertions being re-ordered and Gillian-C used to incorrectly die because of ALoc mismatches. I wonder if there's a way of avoiding this branch explosion though :/ Overall sorting still seems to give better perf in most cases