ESQL: Add exponential histogram CSV Tests

[JonasKunz]

Adds a very basic CSV test for exponential histograms.
The dataset consists of some pathological dummy values alongside with a few real response time histograms.

Adapts the CSV test infrastructure to work with exponential histograms and hopefully doesn't break any bwc / release tests. Please take a closer look at that, as I'm really new to the ESQL testing world.

We don't have a ES test node feature for support of the exponential_histogram field type in elasticsearch, as that is hidden behind a feature flag aswell. But I think using the ES|QL capability should be a good enough surrogate, as we are planning on lifting those at the same time (or at least closely one after another).

[elasticsearchmachine]

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

[nik9000]

Looks fine.

JSON in these is pretty gnarly. For date_ranges we're doing the rust-like 2025-01-01..2026-01-01. We'll do that for ip and numeric ranges too. But these complex things already have json parsing code and it's compelling to reuse it. A local minima, maybe.

Maybe we should do a """ to support un-escaped ". It'd be easier to embed json, I guess.

[JonasKunz]

JSON in these is pretty gnarly

Indeed, but I think querying raw exponential histograms will rarely happen in the CSV tests. We'll rather look at the aggregations (e.g. PERCENTILE) instead.

Maybe we should do a """ to support un-escaped ". It'd be easier to embed json, I guess.

Is this some CSV standard or are you thinking of a custom implementation? I'm only aware of the "" to generate single quotes.

We could also make our CSV-test matching more lenient and omit the quotes completely IINM.