Precisely compute committee weight in FCR

[mkalinin]

Motivation

To calculate committee weights between slots, the existing algorithm relies on an estimate whose accuracy depends on the level of stake consolidation. This PR replaces that estimate with an exact computation, making this part of the algorithm independent of stake consolidation.

Details

• The estimation logic is replaced with obtaining an exact set of validator indices participating in the committees of a span of slots and then computing the total balance of this set.
  Note that if a span of slots crosses an epoch boundary there can be an overlap in participating indices but the final participants set must have no duplicates. Thus, caching each committee weight and using a sum of committee weights as a result will not work in general case.
• This change affects is_one_confirmed and compute_honest_ffg_support_for_current_target which use different balance sources passed to the compute_committee_weight_between_slots.

Tests

• Fixes tests affected by this change, some of the tests are changed significantly as precise computation doesn’t allow for the support discount to reduce the safety threshold to 0; some of the old tests were superseded by new ones
• New tests covering FFG and is_on_confirmed checks are written with the help of AI but reviewed manually

[jtraglia]

Nice, LGTM. I pushed a little nit fix, hope that's fine.

[jtraglia]

Hey @mkalinin feel free to merge this whenever you think it's ready. I'm not sure if you want to have more reviews from others or if there's any reason to wait. Note that you'll need to click the "update branch" button & wait for the CI checks to pass, then you should be able to merge it.