Add cardinality estimate stat

[a10y]

Useful for compressor to decide if Dict compression is worthwhile.

There's a Rust crate already implementing it: https://docs.rs/hyperloglogplus/latest/hyperloglogplus/struct.HyperLogLogPlus.html

Can be used:

• At compress time: determine if Dict is worth trying or just fallback directly to FSST
• At query time: Datafusion allows reporting cardinality estimates, which are used for join selection: https://github.com/apache/datafusion/blob/8ba6732af5f4f32cbe0a23ef6bc2f393c640898b/datafusion/physical-plan/src/joins/utils.rs#L905

[lwwmanning]

I think we want this instead of HLL++: https://www.cidrdb.org/cidr2019/papers/p23-freitag-cidr19.pdf

[lwwmanning]

(In particular, it gives good estimates of cardinality of arbitrary combinations of attributes rather than just attributes, which is cool / handy for compound join keys)

[lwwmanning]

if we're taking off the shelf, this crate looks potentially better: https://github.com/cloudflare/cardinality-estimator/tree/main

[robert3005]

previously mentioned in #85