Improve the parrallelization

Two parallelization scheme are possible :

• by OTB app
• by input file

The first one manage the processing of one file on the CPU and it well adapted to run the chain on local PC. The second one allow to run in parallel several S1 products and finally reduce each projected data to a single S2 tile.

The first step is to check and improve the parallelisation in OTB app. The second step is to separate the chain in one sub-chain which can be run with one S1 product and another one which can merge and the projected data in one S1 tile. The localization of the multi-temporal speckle filtering must be evaluate.

Additional framework (dask for exemple) can be evaluated in a another issue.