New implementation for block iterations and block operators

[tlunet]

Hi @JensHahne,

first of all : during the revision of our convergence analysis paper with Daniel, Robert and Martin, we just realized that some convention notations needed to be improved. I've tried to write down our conventions for this project, in accordance to those in the paper (there is some minor modification, in particular for the index of the block coefficients, see here).

Then, the big change 😄 : I've reworked all block iteration related classes into only two classes, the BlockOperator and BlockIteration, see their base description here. The whole idea is

• to simplify the implementation by reducing the number of required classes and methods
• to allow the definition of a block iteration using string formula (see this basic tuto, and I've also added the new version run in main.py)
• to prepare for the use of matrix for each operator (still have to implement the possibility for an operator to apply the inverse of a matrix ...)

I did not rearranged and cleaned the code yet, but you may see one of the idea behind it : regroup classes in some modules (like block.py or utils.py), where each modules contains the classes or functions for a given functionality, and in a near future regroup all of this in a python package (folder), which name will be the one for the package (not necessarily time4apint, a bit long, I was thinking maybe to blockops, but could be anything else 😉)

Also, don't hesitate to use Jupyter notebooks to create some application examples (for instance, with SDC tests, etc ...). GitHub render those quite nicely, and it won't be too complicated to reuse them for some tutorials for later.

I let you look at this quietly on your own, and we'll see later for the code reorganization ...

Have fun !

PS : also, don't hesitate to correct typo and grammatical mistakes, I tend to make way too much of those ...