fix(v4): MySQL/MSSQL migration bugs found

[dimitri]

MySQL:

• Create target schemas before sequences, ENUM types and tables. ENUM types were created before their schema, so the table using them was lost, along with its FKs and sequence reset.
• AUTO_INCREMENT integer columns become serial/bigserial (mirroring the CL default cast rules), instead of plain integer columns with no sequence.
• DECODING TABLE NAMES MATCHING … AS now reads text columns as CAST(… AS BINARY) and decodes the stored bytes with that charset. SET NAMES asked the server to convert, which double-encoded the very data the clause is meant to repair.

PostgreSQL target:

• Fix the sequence reset off-by-one: setval(seq, MAX(col), true), or setval(seq, 1, false) on an empty table, so the first insert no longer collides with the last migrated row.
• "Reset Sequences" counts sequences actually reset, not tables visited.

SQL Server:

• DEFAULT (N'new') lands as 'new', not as the literal text N'new'; doubled quotes in string literal defaults are unescaped.
• Translate T-SQL default functions: GETDATE(), GETUTCDATE(), SYSDATETIME(), SYSUTCDATETIME(), SYSDATETIMEOFFSET() → CURRENT_TIMESTAMP, NEWID(), NEWSEQUENTIALID() → gen_random_uuid().

Logging:

• Mask passwords in the Source/Target/Connected log lines, for pgloader URIs (user:@) and JDBC URLs (password=).

Tests: unit tests for each fix; E2E coverage in the mysql suite (no more BEFORE LOAD DO create schema workaround, sequences, decoding) and a v4-only tsqldefaults database in the mssql suite. Baselines that recorded the missing nextval() defaults are updated.