There are two parameter models in Npgsql - ordinal and nominal; we use nominal, but the ordinal API is much more efficient; we could ingest nominal const sql and rewrite it as ordinal

this should include splitting semi-colon multi-statements into DbBatch when possible

tasks:

• basic generalized SQL parser
• SQL syntax tests for postgresql peculiarities
• basic exploration of batch concept
• "real" batch concept (UnifiedBatch, should work with/without DbBatch API)
• codegen (including turn-off-and-onable)
• benchmark:
  • vanilla Dapper, parameterized single batch (Npgsql param rewrite)
  • vanilla Dapper, parameterized multi-batch (Npgsql param rewrite+split)
  • AOT Dapper, rewrite enabled, parameterized single batch, DbBatch disabled
  • AOT Dapper, rewrite enabled, parameterized multi batch, DbBatch disabled
  • AOT Dapper, rewrite enabled, parameterized single batch, DbBatch enabled
  • AOT Dapper, rewrite enabled, parameterized multi batch, DbBatch enabled
• use parser in place of the regex in the regular code path? as an option with fallback?
• docs
• performance tuning of parser (I have a plan, don't ask; kind of a rewrite of the parser) (note this is not critical path)