Use CPU and RAM requirements specified by the user for dask workers

[Alputer]

Currently, dask cluster is built using a fixed autoscaler and worker blueprints. We need to specify 2 parameters:

1. CPU and RAM capacity of the workers
2. Maximum amount of workers in dask autoscaler

Also, we need to decide on how we share the configuration between these 2 options.

1. We can have heavy workers with fewer number of maximum amount of dask workers
   --- OR ----
2. Light workers with a higher number of maximum amount of dask workers

We are also facing a possibly major problem in this task, because dask autoscaler does not seem to obey the maximum number of workers rule. In a few compute-intensive tasks, I have observed it exceeds the limit not only by a few workers but even more. If this issue cannot be fixed at all, we might even need to stop using dask autoscaler completely or implement our patched auto-scaling logic.

[Alputer]

This issue is already included in the first set of dask integration pr set whose links are given below. Therefore closing the issue.

#821
reanahub/reana-server#701
reanahub/reana-workflow-controller#600
reanahub/reana-job-controller#464
reanahub/reana-commons#467