Revise bag op RFC-0007 based on latest discussions

[alancai98]

From some recent discussions and changes to the reference implementation + conformance tests, we should revise the bag operator RFC. Relevant PRs / issues:

• PLK reference implementation changes [v1] Refactor representation of set ops partiql-lang-kotlin#1538
• conformance test change -- Fixes an outer union coercion test partiql-tests#120
• conformance test change -- Fix some more SQL set op + PartiQL bag op tests partiql-tests#121
• related spec issue -- RFC-0007 Bag Operator Coercion Behavior Inconsistencies #76

Specifically, the following things may need changed and/or clarified:

1. Coercion behavior in permissive vs strict mode (RFC-0007 Bag Operator Coercion Behavior Inconsistencies #76). Decided we want to error in strict mode when a SQL set op or PartiQL bag op is not a collection; apply coercion function in permissive mode.
2. Coercion of null and missing argument to PartiQL bag op in permissive mode should apply coercion function defined in PartiQL spec related to FROM source coercion. We should coerce null or missing argument to a singleton bag in permissive mode rather than an empty bag.
3. Allow LIMIT,OFFSET, and ORDER BY clauses to apply to SQL set ops and PartiQL bag ops.
4. Clarify treatment of null and missing by providing an example. RFC-0007 states

For all operators, member equality is consistent with the GROUP BY clause.

Would be helpful to clarify if the resultant value of the bag op will also coerce missing to null as we currently do for GROUP BY.