Avoid calling getPivotValues() to improve PIVOT query performance

[labkey-adam]

Rationale

In the absence of an explicit PIVOT ... IN clause, LabKey must execute portions of PIVOT queries to determine the resulting column list. Depending on the query, this can be very expensive; see https://www.labkey.org/OHSU/Support%20Tickets/issues-details.view?issueId=46881. This PR attempts to avoid calling getPivotValues() (the column-list query) unless it's absolutely required. This can improve performance significantly for cases that don't need the query results, e.g., rendering the schema browser tree and populating the linked schema table drop-down list. Approach:

• AbstractTableInfo: introduce initializeColumns(), called immediately before columns are accessed for the first time. Adding columns in this method (instead of the constructor) avoids the potentially expensive operation in cases where the columns aren't used. Call ensureInitialColumnsAreAdded() from all methods that interact directly with _columnMap.
• PivotTableInfo: populate the initial columns in initializeColumns() and defer validation (missing parameter values, etc.) to getAllColumns(). Rework the exception handling in getPivotValues() and its callers to ensure reasonably friendly error messages.

Schema browser tree now avoids executing PIVOT queries. GetQueriesAction (used in linked schema admin page) still indirectly invokes getPivotValues() (e.g., for URL handling). We should either parameterize GetQueriesAction further (includeUrls parameter, maybe false by default?) or switch linked schema admin page to call getSchemaTree.api instead of getQueries.api, to avoid the expensive query. Will address this in a follow-on PR.