Implement approx_distinct using HyperLogLog

[jimexist]

Is your feature request related to a problem or challenge? Please describe what you are trying to do.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]
(This section helps Arrow developers understand the context and why for this feature, in addition to the what)

Implement HyperLogLog functions

Describe the solution you'd like
A clear and concise description of what you want to happen.

Similar to that of presto/trino

Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.

Additional context
Add any other context or screenshots about the feature request here.

[jimexist]

@houqp @alamb is this of any use to your usecases?

[houqp]

It looks like a fun engineering challenge and a really cool feature to add :D I don't have a use-case for this at the moment but I think it's something we would need to support eventually in ballista to close the gap with spark.

[jimexist]

relevant links:

1. smhasher test suite
2. trino uses xxhash 64 bit version
3. hyperloglog.c from redis, which uses murmur3 64 bit version
4. xxhash v.s. murmur3, as of 2016 seems like xxhash has better throughput for larger data size
5. https://github.com/shepmaster/twox-hash
6. https://docs.rs/pdatastructs/0.6.0/pdatastructs/hyperloglog/struct.HyperLogLog.html and https://docs.rs/hyperloglogplus/0.4.0/hyperloglogplus/struct.HyperLogLogPlus.html

[alamb]

@houqp @alamb is this of any use to your usecases?

Yes, I think it is relevant @jimexist and I think it is a great idea. Will try and review #1087 shortly.

Another thing that would be awesome (as a follow on PR) would be to support approx_distinct using Statistics, even when those statistics weren't super accurate (e.g in IOx we have deferred deletes and upserts that are not resolved until actual read time)