Add one out forecast-implied inference in network inference bundle

[guilherme-brandao]

Purpose of Changes and their Description

This PR adds a new feature to calculate forecast-implied inferences when each inferer is removed from the network. These values are critical for evaluating the influence of individual inferers on forecasters' implied predictions.

• Added GetOneOutInfererForecastImpliedInferences function to calculate what each forecaster's implied inference would be when each inferer is removed
• Updated CalcNetworkInferences to include these values in the network inference bundle
• Used existing OneOutInfererForecasterValues type to maintain consistent structure with other one-out calculations

Performance Related Changes:

• The GetNetworkInferences function employs an in-memory cache that temporarily stores results of expensive mathematical operations (in this case, just Exp) to avoid redundant calculations during network inference processing; the cache is enabled at the function's start, utilized during calculations, and then intentionally disabled and cleared upon completion.
• Other minor optimizations.

Link(s) to Ticket(s) or Issue(s) resolved by this PR

Are these changes tested and documented?

• If tested, please describe how. If not, why tests are not needed.
  ✅ Added new unit tests
• If documented, please describe where. If not, describe why docs are not needed. -- Added Data Submission section to ./README.md.
• Added to Unreleased section of CHANGELOG.md?

[xmariachi]

This contains a triple-nested for.
I am concerned about the time this will take when executing in an endBlocker.
Could we time-measure this by having a full max inferers and max forecasters test?
If this has an impact we want to know ... it could add some non-negligible time to the block
cc @spooktheducks @amimart

[amimart]

it could add some non-negligible time to the block

It seems to be used on queries only so it should not impact block production

[amimart]

I've let some time/alloc optimization remarks to see if we can improve it

[amimart]

Left some comments that may be great to discuss

[amimart]

LGTM 👍

[xmariachi]

it could add some non-negligible time to the block

It seems to be used on queries only so it should not impact block production

I think this was left unreplied, but just for completeness: this is internalized into endblocker in another PR