Support for "jittered" delayed retries

[lailabougria]

Through delayed retries, we currently support the exponential backoff strategy, allowing retries to progressively be delayed for longer intervals. However, in high-throughput scenarios, it's beneficial to add jitter to the delayed retry strategy, to prevent retries from causing further spikes of load to already struggling (sub)systems.

Related information

• Exponential backoff and jitter, AWS
• Jitter support in Polly

[ramonsmits]

Similar issue was raised a long time ago:

• https://github.com/Particular/NServiceBus/issues/5012

Implementation:

• https://github.com/Particular/EndToEnd/blob/master/src/PerformanceTests/Common/ErrorProfile.cs

[ramonsmits]

This can have help in "spreading" massive amounts of scatter/gather response processing failures caused by optimistic concurrency control failures and make "waves" of retries "longer" thus reduce failures.