Blocking nvme disconnect fixes

[tiagolobocastro]

fix(agents/ha/cluster): increase path replacement timeout

In a system test run we've seen the disconnect of an old path take 50s
to complete.
We're not sure why this is the case as we cannot reproduce it yet but
for now we can increase the timeout.
This is probably fine as if we can't talk to the node where the app is
then we can't do much more anyway.

We should also consider whether we need to wait for the disconnect at
all. If we can find out if the disconnect has been started, then we
can probably just complete that async and succeed the replace!

Signed-off-by: Tiago Castro <tiagolobocastro@gmail.com>

---

fix(agents/ha/node): disconnect controller async

Dependent nvmeadm crate is blocking and certain operations such as
disconnecting the controller can take quite some time.
Here we are just doing a WA by running the disconnect in a tokio
blocking spawn.
todo: Ideally we should either make nvmeadm async or somehow provide
async as alternative, which won't be great as the code will
likely get duplicated...

Signed-off-by: Tiago Castro <tiagolobocastro@gmail.com>

[tiagolobocastro]

bors merge

[bors-openebs-mayastor]

Build succeeded:

• continuous-integration/jenkins/branch