Add additional checks to EvalEvalBot

[nelaturuharsha]

Add following features to the EvalEvalBot:

• Enforcing score to be between max and min specified. In case min_score, max_score and score_type are not present - the user should be warned.
• Enforcing counts to be integers
• For a given benchmark, the min_score, max_score and score_type should be the same across all occurrences.

[imcoza]

Hey @nelaturuharsha quick question before I start on this
For the "enforcing counts to be integers" check I want to make sure I'm targeting the right fields. Looking at the current schema, the obvious count-like fields (num_samples, num_bootstrap_samples, samples_number) are already declared as int in the Pydantic models, so they'd already fail if someone passed a string or a letter.

The one edge case that slips through is something like 5.0 a float that looks like a whole number. Pydantic silently coerces that to 5 without complaining today.

So just to confirm: are we trying to catch that specific case (reject 5.0 even though it's technically a whole number), or are there other fields you had in mind that aren't covered yet?

Happy to start on the score range and benchmark consistency checks in the meantime since those are clear — just want to get this one right before I touch it.