- Removed duplicated code.
- Made use of qualified names for cf_cariables.
This hides in IndexConfig the expected order of input variable.
- Simplified output type of funs.
No need to create additional types.

It was messing with fully qualified named making python being lost,
and it couldn't choose correctly between the package and the module.

For now, we store the yaml file internally.
This is based on version 0.3.0 of the yaml file.
We must follow updates from https://github.com/clix-meta/clix-meta

This qualifier metadata is used internally to figure out
the necessary parameters to compute the index. We need
it for C3S auto-generated functions.

This module will be used by C3S to
extract icclim indices into individual
functions.

These changes are related to c3s integration.

`transfer_limit_Mbytes` was used to control dask chunking.
This is however not really a good practice because all other
dask configurations are still left outside icclim.

For project related reason this is not a
5.1.0 even if there are many changes since
the initial 5.0 bump.
This should not be an issue as icclim-v5 is
not widely distributed yet.

…_funs

- Overlapping years were not properly computed.
- Chunking was done at dataset level but, it is only needed at data-array level.

This function list all available ecad indices.

This could be improved by destructuring
`user_index` parameter into multiple
parameters on the generated function.

All changes are now under 5.0.0rc3, which
is the targeted official release.

When the in_base and out_of_base
fully overlap, it is unnecessary
(and costly) to bootstrap percentiles.

wip:
- fill the todo
- re-read

…_funs

Also reformulate some sentences.