Field keystone to non dynmic for casting

[pmonson711]

Basic approach, revert to the previous field lookups for realized fields in comparisons. Only use the dynamic value lookup when utilizing a fragment.

TODO TADONE

• I think this means I should rename the field_or_alias to field_or_dynamic, thoughts?
• I'm not sure I like the tuple structure after the update, it really needs a more consistent either type to be more readable and maintainable. Do I refactor those calls now?
• Update the tests for null first / last to have a schema that can have nulls to show the behaviour is correct and to get coverage.
  After some review and thought, the todo approach changed slightly.

1. we switched to maybe_expression instead of field_or_dynamic, and split the calls in fob by the two distinct branches of field and expression.
2. The above addresses this mildly annoying function structure.
3. Added tests for null first and last on day of week date sorting.

[the-mikedavis]

Only use the dynamic value lookup when utilizing a fragment.

seems like a reasonable approach

I think field_or_dynamic sounds good 👍

I don't have strong feelings about a refactor now. The tuple is a bit messy but it's still readable imo.

[pmonson711]

See #11 for some details as we updated the implementation

[the-mikedavis]

Nice. Seems like a good approach to allow the old behavior where we use a field/2 to explicitly list the column when possible and use the fragment/dynamic approach when a field/2 just won't cut it.

Also the wording around "route_" sgtm 👍

[the-mikedavis]

I like 👍