Proposal: Migrate parser library in Spanner support from spansql to memefish

[apstndb]

Is your feature request related to a problem? Please describe.

It seems that spansql package which is used in Spanner support is not actively developed.
googleapis/google-cloud-go#11115

Many of the issues opened due to spansql's lack of support for Spanner SQL syntax.

• Support Default on Google Cloud Spanner #918
• Dropping spanner tables which has row deletion policy #844
• spanner CHANGE STREAM #775
• Module update needed (Cloud Spanner parsing issue) #767
• JSON field type rejected #710

(Although I think some issues are already resolved in latest spansql.)

Describe the solution you'd like

Migrating to another Spanner SQL parser implementation is a solution.
I recommend to consider cloudspannerecosystem/memefish.
Whose Spanner SQL coverage is already superset of spansql.

• Pass all SQL test in spansql tests cloudspannerecosystem/memefish#115

Describe alternatives you've considered

Current usage of spansql is only splitting DDL statements and strip comments.

migrate/database/spanner/spanner.go

Lines 343 to 347 in c378583

	func cleanStatements(migration []byte) ([]string, error) {
	// The Spanner GCP backend does not yet support comments for the UpdateDatabaseDdl RPC
	// (see https://issuetracker.google.com/issues/159730604) we use
	// spansql to parse the DDL and output valid stamements without comments
	ddl, err := spansql.ParseDDL("", string(migration))

I think it can be implemented without full syntax parsing using only a lexer.