feat: EXPOSED-45 Support single statement UPSERT

[bog-walk]

Add functionality for INSERT or UPDATE command.

MySQL, PostgreSQL, and SQLite use their own specific implementations, while the other dialects use the standard MERGE USING statement with aliases and a derived table.

Accepts optional user-defined key columns to determine the conflict condition. If none are provided, the primary key is used. If there is no primary key, the first unique index is used. If there are no unique indices, each dialect handles differently, but any documentation should probably strongly advise that keys are defined to avoid unexpected results.

Accepts an optional WHERE clause, as well as user-defined update assignments if update behaviour differs from insert behaviour. Column expressions, defaults, and subqueries are also enabled if the individual dialect allows them.

Current caveats:

1. Dialects that use MERGE require that the key used in the ON clause is present in the derived table. So if an IntIdTable is used e.g., usually the user doesn't have to manually specify the next id, but upsert will fail unless the id is provided.

Vendor-specific features not currently covered:

• [SQLite] Should be able to chain multiple On CONFLICT causes together.
• [SQLite, PostgreSQL] User-defined update expressions & WHERE clause should be able to access insert values using excluded. identifier. Workaround is to reference the value that would have been inserted directly.
• [Oracle, SQLServer, H2] Technically do allow WHERE clause, but multiple and specific to either WHEN [NOT] MATCH THEN clause (Oracle differs in placement too).

Plan:

1. Submit a second PR for batch upsert, unless advised to push them together.
2. Submit a PR that addresses possible overlap with current implementation of REPLACE (at least in PostgreSQL) & update API docs (the 2 queries are very similar but not equivalent).
3. Add WHERE clause functionality to MERGE implementations.

[bog-walk]

A technicality, but I'm not entirely sure about this choice. If anyone has a cleaner idea of how to replace this chunk in the middle of a statement, I'm open to alternatives. The other way I considered was:

return statement.replaceAfter(" USING ", dualTable) + statement.replaceBefore(" ON ", "")

[e5l]

it's ok to make it this way

[e5l]

please add KDoc

[bog-walk]

Done

[e5l]

lgtm

[e5l]

lgtm

[red-avtovo]

Some examples or a doc page would be much appreciated.
This page covers only one operation, while I'm trying to figure out how to refer to the new value that was intended to be inserted