Improve batched Column performance

[mvilliger]

SmartColumns with batched LookupCall already collected all keys and then updated the cell text as soon as the async LookupCall completed. Therefore, there is only one promise for the LookupCall execution returning the texts for all keys.
Unfortunately the SmartColumn then splits this promise into one promise for each cell containing only the value for the specific cell. This leads to thousands of promises on large tables.
These promises are collected in the TableUpdateBuffer and removed as they resolve. This operation was O(n^2) as the buffer internally used an array to store the promises.

Processing lots of such promises together with the inefficient handling of the buffer leads to unnecessary computation time on the client. Because of some async overhead, executing a promise for each cell is significantly slower than executing one promise handling all cells. On large tables with thousands of rows and multiple SmartColumns this may sum up to several seconds just to process these promises.

This commit changes the following:

• Introduce a helper class BatchCall to collect keys and using an async   batch callback executed for all the keys.
• Column uses this new BatchCall in formatValue() if available. This   allows each column to support batch processing more easily.   Implementors only need to set the _batchFormat member defining the   batch callback.
• The deferred Cell text update in batch mode always updates all texts   of the column instead of only one cell. Accordingly, _formatValue and   formatValue() in the async case now return all texts instead of only   the one for a single cell.
• StaticLookupCall & LookupCallColumn now supports batching to benefit   from the new features.

410165, 399939