Increase maxPerformDataSize for maxObservation, maxOutcome size

[infiloop2]

maxPerformDataSize was increased to 10000 from 5000, it breaks current maxObservation and maxOutcome sizes

• For maxObservationSize, I had to reduce the number of allowed upkeeps instead of increasing observation size as libocr puts a limit of 1Mb. This should be ok when new nodes will send less number of upkeeps that old nodes should be able to parse. however on new nodes we do validations on number of performables from old nodes' observations, which will break if it goes above 50. Our current plan is that we don't expect it will be hit in prod during upgrade.

• For maxOutcomeSize, i could increase it to fit the given upkeeps. This should also be ok as it increases max outcome size, old nodes outcome would still be valid. If new nodes outcome size increases 2Mb, then old nodes won't be able to handle it but that was broken anyway

A more robust solution is to calculate this limit dynamically but that is considerably more involved. created auto-7597 for it

Test Plan: Will rely on existing CI tests and node upgrade tests, with known expectation that upgrade will fail if there are more than 50 upkeeps eligible per second during upgrade

[infiloop2]

this is a smaller change than 2x because

1. there was existing margin
2. there is a component of proposals in outcome that didn't scale according to performData