RD-15410: DAS function support

[bgaidioz]

(Goes with raw-labs/multicorn-das#28)

This PR adds DAS function support to multicorn by introducing:

• A new C function entry point multicorn_function_execute, which can be invoked
  from SQL with a set of parameters that includes FDW-like options, named
  argument strings, and the actual function parameters.
• A bridging helper function das_function_execute that:
  • Converts options and arguments into Python dictionaries,
  • Dynamically imports and calls the Python module/class (multicorn_das.DASFunction.execute_static) for DAS function execution,
  • Then converts the Python result back into a PostgreSQL Datum.
• An auxiliary function build_dummy_attinmeta(Oid retType) for creating a
  single-column AttInMetadata structure, used when converting the returned
  Python object to a Datum of the indicated retType. If the return type is an
  array, it sets attndims=1 to handle array‐like results.

A fix:

• INTERVAL was read from a dict representation, that isn't anything standard (it is the Python object generated by multicorn-das), to turn it into an ISO string which Postgres can parse and import. If multicorn-das generates that a Python ISO string instead of that dictionary, then importing an interval is simplified: we can pass the string to Postgres.
• That implementation also helps with extracting fields from a JSONB returned by a DAS function: if the JSONB contains the JSON representation of the original Python dict, a user wouldn't easily translate it into a Postgres INTERVAL, while if it comes as an ISO string, value::interval works.

multicorn_function_execute

Runs the logic to execute a DAS function statically.

• It receives at least two arguments:
  • a text[] of options, which permits to identify which DAS and function to execute,
  • and a text[] of parameter names, which is used eventually to build the named arguments as a Python dictionary,
  • plus N further actual arguments for the DAS function.
• The argument/return types are retrieved from pg_proc, in order to calls foreign_function_execute, that bridges to Python.

foreign_function_execute

It bridges the SQL function to Python:

• Converts FDW options to a Python dict.
• Converts named function arguments to another dict.
• Imports multicorn_das.DASFunction and calls execute_static(...).
• Uses pyobjectToDatum(...) to transform the Python result to a Postgres Datum.

build_dummy_attinmeta

Converting the function result to Datum reuses the existing Python conversion
code that runs for tables. Because that code takes its conversion information
from column descriptions, we add build_dummy_attinmeta, a function that
returns a single‐attribute AttInMetadata with the given type for the required
text-based conversion that are already in place for column values.