Batch processor: strictness support

[jmacd]

In the batch processor we have a bit of logic with the following effect:

When the flush reason is because of size (as opposed to shutdown or timer), AND
when there is a maximum batch size, AND
when there is a minimum batch size, AND
when more than one batch was output, AND
the last batch is less than the minimum size, THEN
withhold the last request, retaining it to accumulate a full batch and resetting the flush timer.

This is a risky operation. It requires a certain degree of correctness in the lower-level otap_df_pdata::otap::batching logic. While it appears that this logic is correct, the consequences of it not being correct are severe, as it will mean acking or nacking incorrect inputs.

Proposing a way to disable this feature, by setting withhold_residual: false (or something like that).