Support pandas/datetime Timedelta/DateOffset as literals converted to SQL intervals

[knassre-bodo]

The goal of the issue is to add the following support for PyDough literals:

• Any datetime.timedelta can be passed in, and is threaded throughout the pipeline until SQLGlot conversion where it becomes an equivalent INTERVAL syntax.
• Same for pandas.Timedelta and pandas.DateOffset with all standard ANSI units: YEAR, MONTH, DAY, HOUR, MINUTE, SECOND (no subsecond, for now, and expand quarters/weeks into months/days).
• Allow passing in these units as modifier arguments to the DATE function, or just directly adding/subtracting to date/datetime arguments.

Will need to add a new PyDough type for intervals/deltas. The generation of SQLGlot code will need to be quite different for SQLite, which means full type inference is required to realize when the addition/subtraction of such an interval is happening so it can be converted to an equivalent DATE expression, and so using an interval in any other manner is unsupported. This means that type inference support (#194) and the DATE-esque functionality (#208) are required prerequisites.