Don't log on replay

[yaythomas]

Context

GitHub Reference: logger.py#L45

There is a logger on DurableContext. Currently the logger will always log, even on replay. Amend this logger so it uses the current execution state to compare existing checkpoints to the current logger.

Implementation Notes

ctx.logger.log('first log')

ctx.step('step1')

ctx.logger.log('2nd log')

ctx.step('step2')

In each context keep track of which operations you have "visited".

During replay, "step 1" should be in the execution state.
If you have "step 1" exists, but you haven't run the ctx.step line that created it, this must be a replay.

Once all operations in a context are crossed, it is no longer replay and executing new code.

This works because all operations must be created linearly (for deterministic step ids), and we require child contexts for concurrency.

Acceptance Criteria

• Only log on replay for the default logger
• Provide a way for custom loggers to take advantage of the same replay checks

Priority: Low
Estimated Effort: 3 days

[leandrodamascena]

Closed via #88