Mark Gorilla codec on non-float columns as suspicous

[rschu1ze]

This PR marks the Gorilla codec on non-float columns as "suspicious codec".

Reasons:

1. The original Gorilla paper proposed a compression schema for pairs of time stamps and double-precision FP values (see "Gorilla: a fast, scalable, in-memory time series database", Sec. 4.1.2, "Gorilla restricts the value element in its tuple to a double floating point type."). ClickHouse's Gorilla codec only implements compression of the latter and it does not impose any data type restrictions.

• Data types != Float* or (U)Int* (e.g. Decimal, Point etc.) are definitely not supposed to be used with Gorilla.
• (U)Int* types are debatable. The paper only considers integers-represented-as-FP-values, a practical use case for which Gorilla works well. Standalone integers are not considered which makes them at least suspicious.

2. Achieve consistency with FPC, another specialized floating-point timeseries codec, which rejects non-float data.

3. On practical datasets, ZSTD is often "good enough" so it should be okay to disincentive non-ZSTD codecs a little bit. If needed, Delta and DoubleDelta codecs are viable alternative for slowly changing (i.e. time-series-like) integer sequences.

This PR disallows the creation of new Gorilla-compressed non-float columns unless setting "allow_suspicious_codecs" is "true". Existing data of this kind will continue to load just fine because setting "allow_suspicous_codecs" is ignored during server startup and in ATTACH TABLE.

Side note: Gorilla-compressed non-float columns should not be very common in practice since the documentation always listed Gorilla as a specialized codec for "series of floating-point values".

Changelog category (leave one):

• Improvement

Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):

Mark Gorilla compression on columns of non-Float* type as suspicious.

[alexey-milovidov]

@rschu1ze IIRC, we allow all codecs (including suspicious) for ATTACH TABLE queries and on server startup. So, why don't turn the check on by default?

[rschu1ze]

@alexey-milovidov Thanks, you are right ... I verified locally that server startup and ATTACH TABLE don't care about suspicious_codecs. So basically, when we make Gorilla-on-nonfloats suspicious right now, the creation of such tables will only work with allow_suspicous_codecs = true but all existing use cases will continue to work. Cool, then I'll change this PR, it will simplify things.

[rschu1ze]

Test failures are all unrelated or fixed in master.

[rschu1ze]

@alexey-milovidov Approve + merge? Thanks!

[alexey-milovidov]

Is it ok for Nullable(Float)? Array(Float)? Tuple(Float, Float)?

[rschu1ze]

Is it ok for Nullable(Float)? Array(Float)? Tuple(Float, Float)?

Fixed that, thanks.