[RFC007] Clean the AST alloc interface

[yannham]

A previous PR introduced generic alloc and alloc_iter methods, instead of needing a myriad of xxx_move functions for each and every AST component type. However, this poses a risk of memory leak because this method allocates in the generic bumpalo arena, which doesn't run destructors.

This commit makes this approach safer by using a marker trait which forbids some AST components from being allocated through this method, when they need to be dropped. Anecdotally, alloc_iter is renamed to alloc_many and a specialized version alloc_singleton is added to eliminate a lot of alloc.alloc_many(iter::once(x)).

In a second time, all the legacy xxx_move variants or the like are removed, and replaced with alloc and alloc_many whenever possible, making the interface of AstAlloc quite smaller.

[github-actions]

Bencher Report

Branch	rfc007/streamline-alloc-interface
Testbed	ubuntu-latest

Click to view all benchmark results

Benchmark	Latency	nanoseconds (ns)
fibonacci 10	📈 view plot 🚷 view threshold	495,240.00
foldl arrays 50	📈 view plot 🚷 view threshold	1,642,500.00
foldl arrays 500	📈 view plot 🚷 view threshold	6,709,200.00
foldr strings 50	📈 view plot 🚷 view threshold	7,053,000.00
foldr strings 500	📈 view plot 🚷 view threshold	62,180,000.00
generate normal 250	📈 view plot 🚷 view threshold	46,545,000.00
generate normal 50	📈 view plot 🚷 view threshold	1,988,300.00
generate normal unchecked 1000	📈 view plot 🚷 view threshold	3,331,500.00
generate normal unchecked 200	📈 view plot 🚷 view threshold	759,280.00
pidigits 100	📈 view plot 🚷 view threshold	3,206,100.00
pipe normal 20	📈 view plot 🚷 view threshold	1,490,500.00
pipe normal 200	📈 view plot 🚷 view threshold	10,258,000.00
product 30	📈 view plot 🚷 view threshold	828,370.00
scalar 10	📈 view plot 🚷 view threshold	1,508,800.00
sum 30	📈 view plot 🚷 view threshold	825,910.00

🐰 View full continuous benchmarking report in Bencher