@copilot Can you create a pull request that targets the "landlock-arraydomain" branch (not main!!) that does the following?

I was talking to the maintainer about converting Landlock to use a flat-array based lookup of rules, rather than the current rbtrees. Here is a snippet:

maybe the optimal way to do it would be to

1. Add a "u32 next_collision" to domain_index
2. Sort the indices array such that each domain_index is either placed in the slot corresponding to a hash of the key, or placed in one of the empty slots if a collision happens (and updating next_collision)

And then we can either allocate the array to be the next power of 2 size, in which case hashing would be faster, or allocate it with exactly the number of indices we need and do a modulo operation (less space, but potentially slower as it has to do a division?)

I can try to implement that. Searching is easy, but construction would probably be more tricky, and would also (temproarily) require a u32[number_of_rules] (or u32 [number_of_slots] if rounding to power of 2) scratch array

Please refer to #9 for the current flat array domains impl. We still want a flat array and we don't want individual allocations for each rule, so don't use stuff like hlist.