retry backoff strategies

[yaythomas]

Currently the retry decorator works with a single static sleep interval.

Add backoff strategies better to enable linear, exponential sleep intervals in between retries with randomization (jitter).

Out of box, provide (assuming sleep interval is x and iteration/retry counter is n)

• fixed (default) - sleep = x.
• list - sleep = [x0, x1, x2] sleep picks the next value from a list. keep on returning last value when reaches the end and more retries necessary.
• linear. sleep = nx
• exponential. sleep = (2**n)*x
• jitter variants for all the above (randomly select interval between a lower and upper bound given by the formula of the strategy)

In line with pypyr's ethos for complete customisation, also allow operator to create their own custom back-off strategy that will dynamically load as do steps/content parsers etc.

Thus,

step: mystep
retry:
  max: 3
  backoff: mypackage.mymod.myretrybackoffstrategy