[ML] Improvements to model memory limit handling for data recognizer modules

[peteharverson]

Meta issue outlining the work that is needed to improve the model memory limit handling for anomaly detection jobs by the data recognizer modules.

• Add a flag to the module setup endpoint to do an estimate of the model memory limit for the job configuration based on actual input cardinality, in place of the hardcoded value specified in the module ([ML] Module setup with dynamic model memory estimation #60656)
• Add an option to the data recognizer wizard for module jobs to use an estimate based on actual input cardinality ([ML] Module setup with dynamic model memory estimation #60656)
• Enable the model memory limit estimate to be done for jobs created from the ML data recognizer job wizard ([ML] Enabling mml estimation in data recognizer module setup #64900)
• Change default behavior of modules setup endpoint to run the model memory limit estimate for all jobs in the module ([ML] Enabling mml estimation in data recognizer module setup #64900)
• Data recognizer to warn if module jobs will be assigned lazily due to lack of resource. ([ML] Lazy ml node UI improvements #90455, 7.11.1)

[elasticmachine]

Pinging @elastic/ml-ui (:ml)

[peteharverson]

Closing as all the items have now been addressed.