Prevent multiple active instances with the same id

[cschleiden]

With this change there can only be one active workflow instance with the same ID. If one attempts to create either a workflow or sub-workflow instance with the same ID a backend.ErrInstanceAlreadyExists error will be returned.

MySql and SQLite backends both make an additional query when creating a new workflow or sub-workflow instance.

The redis change is a bit bigger: checking for top-level instances with the same ID is relatively easy, but checking for existing subworkflow instances within the larger CompleteWorkflowTask transaction is not 100% safe. In practice it would probably be fine, preventing sub-workflow id collisions is relatively easy but I decided to try to move the full CompleteWorkflowTask transaction to a .lua script instead. That way no other commands are executed while this is being active.

The script requires quite a bit of data, so every potential key needed and every argument needs to be passed into from the calling Go code. In the script a couple helper functions getKey() and getArgv() return the next key/argument. This requires careful matching between Go and script and might be optimized by passing in a json struct instead that is then unmarshaled via cjson in the script but this works for now.

[cschleiden]

!bench !large !skipmysql !skipsqlite

[cschleiden]

!bench !large !skipmysql !skipsqlite

[github-actions]

Redis run

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
redis-main	161.4 ± 19.3	124.4	194.2	1.15 ± 0.19
redis-pr	140.2 ± 16.4	112.3	172.0	1.00

Large Redis payload run (1MB)

Command	Mean [s]	Min [s]	Max [s]	Relative
redis-main	21.719 ± 0.625	20.614	22.596	1.00 ± 0.04
redis-pr	21.628 ± 0.737	20.663	22.775	1.00