hoon: make sane memory-efficient for large atoms

[johnhyde]

Description

Resolves #6380.

I updated the section of sane which checks for %t validity to iterate through bytes with an incrementing cursor. This greatly reduces memory load.

[joemfb]

This is a good catch! +rsh loops are convenient to write, but horribly inefficient.

I have a style nit:

[joemfb]

The indentation here is not idiomatic, it should look something like:

=+  tef=(teff cur)
?&  ?|  =(1 tef)
        =+  i=1
        |-  ^-  ?
        ?|  =(i tef)
            ?&  (gte (cut 3 [(add i inx) 1] b) 128)
                $(i +(i))
    ==  ==  ==
  $(inx +(inx))
==

[johnhyde]

Ok, thanks. I was getting mixed up with the indent style for ?- and ?+. But shouldn't $(inx +(inx)) be at the same indentation level as the == == ==?

[joemfb]

Yeah, you're right.

[johnhyde]

Can we squash these commits into a single nice message on merge, or do I need to try to rewrite history to abide by the commit message guidelines?

[joemfb]

PRs are usually merged instead of squashed, so you can use rebase to squash these commits down to one.

this will do

[jalehman]

We normally don't squash commits, so this is good to go with no further changes needed.