[UDFs] Ability to specify a function's `volatility`

[QMalcolm]

Housekeeping

• I am a maintainer of dbt-core

Short description

Data warehouses have a general concept of volatility when defining UDFs. Defining a function's volatility lets the data warehouse do some optimizations when executing the function.

• Snowflake uses VOLATILE / IMMUTABLE (default is VOLATILE)[docs]
• Redshift uses VOLATILE / STABLE / IMMUTABLE (default is VOLATILE)[docs]
• BigQuery does not support this (it infers it as best it can)
• databricks use DETERMINISTIC (assumed nondeterministic unless declared)[docs]

Acceptance criteria

• A volatility can be declared in a function's config
• By default the volatility of a function is None
• A volatility can be defined as deterministic, stable, non-deterministic, or None`

These next two should be moved the dbt-adapters related issue once it exists

• If a volatility is specified for a function, then the create function statement run against the data warehouse uses the appropriate volatility keyword
• If volatility is None for a function, then no volatility keyword is used in the create function statement run against the data warehouse.

Suggested Tests

Can specify a volatility

Impact to Other Teams

N/A

Will backports be required?

No

Context

No response

[graciegoheen]

@QMalcolm Thoughts on making this a property instead of a config?