Add retrying to the dotnet ArmadaClient

[robertdavidsmith]

All our users have to implement retrying themselves.

However most don't implement it until there is connection problems (currently caused by load balancers).

Ideally ArmadaClient would come with some default retrying (or we could provide a RetryingArmadaClient in the library and users can choose)

This saves users from:

• Having to implement it themselves for basic retrying

• Reduces the amount of woe users experience due to occasional network issues

@JamesMurkin has full context, suggest getting in touch with him before starting work

[robertdavidsmith]

James Murkin: Some cases we'd like to be able cover with retrying:

• Extended outage (5+ min). Ideally watch would keep retrying for a while and when a load balancer dies it kills many clients connections. Some handle this gracefully, some require the user to rerun the jobs

• Minor network blips <1minute