Add several fixes and improvements in Ekka and Mria.

Ekka:
  - improve cluster discovery log messages to consistently describe actual events
    [Ekka PR](https://github.com/emqx/ekka/pull/204)
  - remove deprecated cluster auto-clean configuration parameter (it has been moved to Mria)
    [Ekka PR](https://github.com/emqx/ekka/pull/203)

Mria:
  - ping only running replicant nodes. Previously, `mria_lb` was trying to ping both stopped and running
    replicant nodes, which might result in timeout errors.
    [Mria PR](https://github.com/emqx/mria/pull/146)
  - use `null_copies` storage when copying `$mria_rlog_sync` table.
    This fix has no effect on EMQX for now, as `$mria_rlog_sync` is only used in `mria:sync_transaction/2,3,4`,
    which is not utilized by EMQX.
    [Mria PR](https://github.com/emqx/mria/pull/144)