Improve SQL Server migration process

[labkey-adam]

Rationale

We need to lowercase all ENTITYIDs during SQL Server migration. Also need to address a few issues and complications introduced by the cloning migration work and other recent changes.

Fix a variety of SimpleFilter issues uncovered during the cloning migration work. Add regression tests for these issues.

IN (NULL) has been used in various places as a substitute for FALSE. This construct is problematic because NULL is considered an unknown value in SQL, which means IN (NULL) evaluates to FALSE but NOT IN (NULL) also evaluates to FALSE, which can lead to surprising results. For example, in LabKey, new InClause(fk, emptyList) selects no rows and new NotClause(new InClause(fk, emptyList)) also selects no rows.

SQL Server Migration Changes

• Eliminate DatabaseMigrationConfiguration.getColumnNameFilter(); each configuration's getTableSelector() method can filter columns, if needed.
• Bootstrap Users and Guests groups during bootstrap. Delete them before migration.
• Rework attachment copying to simplify the SQL Server migration case: all rows are copied with no filtering and no need to enumerate or track AttachmentParentTypes. Recognize Image data type as a large binary type.
• Provide a mechanism for schemas to map arbitrary GUID values that reside in non-GUID-typed columns to lowercase. Implement for core.PortalWebparts (GUID values are embedded in webpart properties) and comm.Announcements (GUID values can be used as DiscussionSrcIdentifiers).

PostgreSQL Dialect and SimpleFilter.FilterClause Changes

• BasePostgreSqlDialect is common to Redshift and PostgreSQL; move PostgreSQL-specific methods from BasePostgreSqlDialect to PostgreSql92Dialect.
• Fix SQLClause.equals()
• Fix SqlDialect.testLikeOperator() when scopes with different dialects are present. All configured scopes are tested, but the debug string used by the test was always generated assuming the LabKey scope.
• Replace use of IN (NULL) to represent FALSE with 0 = 1