ESQL QL Grand Refactor

[nik9000]

Description

When we built ESQL we built it on top of QL's core, which itself was refactored fairly quickly out of SQL. It's what we had time for and it had a lot of goodies. But it brought a lot of baggage. SQL and EQL live on top of _search and ESQL doesn't, for example.

• Make a copy of the ql project and make esql depend only on this.
• Find all of the tests in SQL, EQL, and QL that we implicitly rely on from ESQL and port them to ESQL (at least optimizer rules tests + expression translator tests)
• refactor/trim the Expression tree
• Unify data types management (ie. avoid mix ofDataTypes and EsqlDataTypes)
• Unify type resolution management (ie. avoid mix of TypeResolutions and EsqlTypeResolutions)
• Replace EvalMapper with calls to EvaluatorMapper#toEvaluator
• refactor/trim the ExpressionTranslators
• Move everything from EvaluatorMapper to ESQL's Expression superclass
• Double-check exception management (ie. make sure that all the exceptions thrown are EsqlClientException, and not only QlClientException)
• Double-check the usage of QL optimization rules (see QL OptimizerRules vs ESQL OptimizerRules)
• Have our own LogicalPlans and PhysicalPlans

[elasticsearchmachine]

Pinging @elastic/es-analytical-engine (Team:Analytics)

[nik9000]

• Double-check exception management (ie. make sure that all the exceptions thrown are EsqlClientException, and not only QlClientException)

The rest of Elasticsearch doesn't really use special exceptions for a project. I think it'd be worth seeing if we should standardize on the assortment of exceptions the rest of Elasticsearch uses. Not sure we'd actually do it, but worth checking.