minor optimization: avoid adding new values rows on del in slices

[pascallouisperez]

Currently, we store refs to slices in the worksheet_values table as [:max_rank:slice_id. Since we always add a row if a value is dirty, we "duplicate" rows with same max_rank on deletes.

For instance, here is an example of three consecutive changes: 1. add an element, 2. add a second element, and 3. delete both elements. The resulting worksheet_values table looks as so:

ws_test=# select from_version, to_version,value from worksheet_values where index = 42;
 from_version | to_version |                   value                   
--------------+------------+-------------------------------------------
            1 |          1 | [:89:ba50ec84-4444-4325-a372-d8c430429fa6
            2 |          2 | [:90:ba50ec84-4444-4325-a372-d8c430429fa6
            3 | 2147483647 | [:90:ba50ec84-4444-4325-a372-d8c430429fa6
(3 rows)

For the 3rd change, we need to update the worksheets table's version column (for concurrency control), as well as update the worksheet_slice_elements to change the to_version of the deleted elements, but we do not need to add a row in the worksheet_values table.