Fix entry indexes.

The NXtomo refactoring a broken broke feature (very important for data automation):
- get / deduce entry name according to the INITI-ACQUISITION ENTRY.
  This was insuring consistency of the created NXtomo and avoiding conflict in naming files and entry.

This commit intend to fix the break done and even improve this mecanism by adding the `bam_single_file` option to inforce the creation of one file per entry.

Now each acquisition has a `start_index` values to know from which index he can create entries and files.
Each acquisition must also provide a way to inform the converter on many entries he will create (at most) from the `get_expected_nx_tomo` function.

For the ZSerieAcquistion and PCOTomoAcquisition this is not possible to provide during discovery of the initialization step.

* The ZSerieAcquisition will increase the start_index for the next acquisition each time he register a new url
* The PCOTomoAcquisition will postpone this when he will parse url for Projection. where this information is registered

It would be beniefic to refactor the `build_acquisition_classes_frm_titles` and `build_acquisition_classes_frm_urls` functions.
4 jobs for fix_entry_index_deduce_n_nxtomo in 3 minutes and 4 seconds (queued for 3 seconds)