Risk calculation in solvers crate

[sunce86]

Description

Implements #1899

Changes

Risk parameters are now part of solvers crate configuration.
Based on risk parameters, success_probability is calculated and forwarded to driver.

This PR is aware of bad gas estimates used. Better gas estimates might come from #1531
Currently, Dex and Baseline and Naive solvers are using these bad gas estimates

Release

Risk parameters need to be defined in the configuration. Risk parameters should be extracted from https://github.com/cowprotocol/services/blob/main/crates/solver/src/solver/risk_computation.rs#L53

IF we want to stop using bad gas estimates and just fallback to success_probability=1, then risk parameters need to be set as [0,0,0,f64::MAX].

[fleupold]

while Naive solver returns success_probability=1 since it does not have any gas estimate.

The naive solver always operates on a single liquidity instance, which has a gas estimate attached to it (requires adding the constant settlement overhead). Can we add this logic to the PR (should be quite small)?

[nlordell]

LGTM. Approving assuming:

• The tests::Config::None variant is removed if it is indeed no longer used
• @fleupold's comment of getting the gas estimate for the naive solver is implemented (should be doable in the same/similar way to the Baseline solver)

[fleupold]

LGTM, just an issue with the gas math imo.