Add regr_slope() aggregate function

[2010YOUY01]

Which issue does this PR close?

NA

Rationale for this change

I'm planning to add PostgreSQL linear regression aggregate functions. Their implementations are similar, so I'd like to implement regr_slope(y, x) first, and the remaining functions should be easy to extend.

What changes are included in this PR?

Adding regr_slope(y, x) aggregate function, mainly implementing Accumulator for RegrSlopeAccumulator
The linear regression is calculated using COVAR_POP(x,y) / VAR_POP(x)
One-pass and parallel algorithm for covar/var reference: https://en.wikipedia.org/wiki/Algorithms_for_calculating_variance#Welford's_online_algorithm

Are these changes tested?

All end-to-end tests produced the same result compared to Postgres

Are there any user-facing changes?

No

[alamb]

Thank you @2010YOUY01 -- I'll check this out carefully over the next few days.

One thing I wonder about is how large the datafusion library is getting. I was thinking maybe it would be useful to start splitting some of the more advanced / special purpose functionality into their own crates, as partly described in #7110. Curious what you think about this

[alamb]

I am sorry I ran out of time today to review this. It is on my list for tomorrow

[alamb]

Thank you @2010YOUY01 -- I found this PR easy to read and well tested and documented. I think it is a very nice contribution

Looks like it is implemented in several other databases:

https://www.vertica.com/docs/9.3.x/HTML/Content/Authoring/SQLReferenceManual/Functions/Aggregate/REGR_SLOPE.htm

https://docs.snowflake.com/en/sql-reference/functions/regr_slope

[alamb]

👍

[alamb]

I don't really have any suggestions or comments for this PR so merging it in