feat: solving auctions tutorial

[mfw78]

Problem

Trying to start to run a solver, or write a solver is already a very difficult problem - and out of date documentation doesn't help 🙃

Suggested solution

Rewrite the "tutorials" section for solvers, also taking into account the new driver/solver colocation architecture.

Alternatives considered

N/A

Additional context

Considerable work has been done consolidating the section from the old documents. Work in this area can be found at https://github.com/cowprotocol/docs-v2/tree/main/docs/cow-protocol/tutorials/solvers.

Acceptance criteria

• Inaccurate information removed from documentation
• Basic tutorial of starting a solver

[KaiserSnipes]

Hey! I run a solver on CoW (Arbitrum and Base), and I honestly ran into a lot of these tutorial gaps myself while getting set up, so if this is still open for contributions I'd be happy to help chip away at it.

A few things I bumped into that seem to line up with the acceptance criteria. A bunch of the pages under cow-protocol/tutorials/solvers/ are marked draft: true and 404 on the live docs right now: create, deploy, evaluate, test, workshop, and from_shadow_to_prod (that last one just redirects to the category index). So the whole "okay, how do I actually try my solver before going to prod" part of the journey is basically missing. The evaluate.md page does technically exist, but it only covers the driver notify endpoint, with nothing on testing offline, reading your shadow results, or figuring out why a settlement simulation failed.

One other small thing that fits AC#1 (removing inaccurate info): onboard.md now tells new CoW-DAO-pool solvers to start on BNB Chain first, but the shadow-competition section a bit further down still says shadow is "only arbitrum and mainnet." So someone following that advice to launch on BNB doesn't actually have a documented way to test before prod on the chain they were just told to start with.

If it'd be useful, what I'd love to contribute is a practical "evaluate your solver" guide from an operator's point of view, basically the pre-prod options that actually exist today. That'd cover a self-hosted offline path (replay recent auctions against your own /solve endpoint and sanity-check limit prices, no-bleed, and EBBO before you touch anything hosted; I've built an open-source backtester that does this, though I'd keep the guide itself tool-agnostic), how to run a self-hosted shadow (autopilot --shadow ) or the playground fork stack against any chain including Base and BNB, and a short troubleshooting section for the usual first-timer pain (sim failures, unfunded EOA / settlement mismatch, limit-price and EBBO rejections), since there aren't any troubleshooting pages at the moment.

I'm happy to start small, maybe the evaluate page plus a troubleshooting page as a first slice, and I'd rather line up on scope with you first so it fits the rewrite you had in mind instead of going off and writing the wrong thing. Let me know if that sounds helpful!