computron limit failed to keep swingset compute time in wall-clock bounds

[dckc]

Describe the bug

escalating from #6625 (comment) :

This is followed by a series of blocks which each use the full 65Mc. The amount of wallclock time they take varies, however, from as little as 4s, to a lot of 20-30s blocks, a pair of 55s, and a single 122s outlier.

To Reproduce

Steps to reproduce the behavior:

1. launch a chain using pismoA software
2. add ~400 smart wallets
3. add vbank asset

Expected behavior

Work should be split among blocks that are roughly 6 seconds, with outliers around 25 sec.

Platform Environment

only observed on mainnet, so far

[raphdev]

We can repurpose the XS fuzzer, using a matching computron limit (65M) and timeout (6s) to get a corpus of test cases that time out but never reach the limit. This may be helpful in investigating where there are costs in accounting (i.e., unmetered or under-metered operations).

I ran a test over a few days and collected 150+ examples of snippets that seem to reproduce this issue.