Design for "hash arrays" needs to clearer

[thejayps]

This issue captures several defects/comments recorded during the review of #230:
"M. https://github.com/Ravenbrook/mps/blob/branch/2023-05-31/hash-arrays/code/poolamc.c#L961-L962 "defer the size accounting" with what intention? (Could refer to design.mps.strategy section.) rule.generic.clear (Please see
#233 (comment) when editing this. RB 2023-05-07)"
from #230 (review)

"My only comment is that the mechanism that setting
MPS_KEY_AP_HASH_ARRAYS to true uses to not contribute to nursery space
is not described. But perhaps it is correct that implementation
details (which could change) are not discussed."
from #230 (review)

"NQ. Is ramping documented? If so, we could mention in the manual that hash arrays are ramping." from #230 (comment)

Also: "comment1: nice improvements to the documentation noted in the past could be relevant https://www.ravenbrook.com/project/mps/issue/job004109/" from
#230 (comment) could be relevant if it still applies. This needs investigating.

[thejayps]

Assigned optional since having up-to-date and clear design for hash arrays will certainly help developers work on hash arrays in future, however the operation of hash arrays is probably simple enough that it could probably be inferred from the code successfully with some additional effort. Improving design for hash arrays would be of benefit and it's worth assigning resources to this.