Templates support

[john-sanchez31]

Linked ticket

Closes #173

Type of change

• Bug fix
• New feature
• Refactor
• Docs / config

What changed and why?

Adds support for PyDough templates — reusable, parametrized PyDough/Python snippets registered in metadata, so common query patterns can be defined once and reused. This PR adds:

• A new templates metadata object, with attributes (labeled values usable as template arguments) and definitions (the templates themselves).
• Support for calling a loaded template directly, as an ordinary PyDough function, or via the new pydough.call_template(name, labels=...) API using human-facing labels.
• Support for recursive templates, templates returning plain literals, and templates that build a user-generated collection from an input pd.DataFrame.

How I tested this?

• New test file test_templates.py, covering end-to-end template calls (direct and via call_template), metadata parsing/validation error cases, and call_template runtime error cases (unknown template, unknown/missing labels, type mismatches, disallowed usage).

Notes for reviewers

Minor from_string refactor as part of this: from_string and call_template now share a common _execute_source helper, with a return_unqualified flag (True requires the source to produce an UnqualifiedNode; False allows any return type, needed for literal-returning templates).