collect all jobs before starting execution

[malteish]

currently, when starting the canister, it might catch up on jobs, resulting in reverts.
Example: we scan 100 blocks, and job a should be executed every 10 blocks. It was executed in blocks 10, 20, 30, 40, resulting in 4 logs that order new executions. The canister would currently try to execute the job 4 times because it does not check if the execution has taken place. Instead of starting a timer each time, it should keep a list like this:
job -> nextExecutionTime

When scanning the 4 logs, nextExecutionTime is updated 4 times without execution. Only when the current block is reached should we scan for which job to execute.