Improve CI pipeline of fetchers

Related to #507 (closed) (maybe do it first)

This issue is to be detailed...

Description

Goals:

• currently CI convert and download jobs are defined by one job
  • we should use 2 different jobs
• performance: avoid cloning Git repositories, especially big ones
  • takes > 45min for "big" fetchers (WB, ECB...)
  • note: we already have a stateful environment for Eurostat and IMF (cf wiki)
• implement locks on pipelines
  • for example to prevent running INSEE download twice in parallel... otherwise git push of one of them will fail...
  • cf https://gitlab.com/gitlab-org/gitlab-foss/issues/20481