[ML] Improve UX regarding df analytics model memory limit

[przemekwitek]

When the user is about to create df analytics job, the DataFrameAnalyticsConfig.model_memory_limit field should be auto-populated if the user did not provide any value for this field.
More specifically:

1. _ml/data_frame/analytics/_estimate_memory_usage API endpoint should be called
2. response.expected_memory_without_disk (this may change in future releases) value should be copied to DataFrameAnalyticsConfig.model_memory_limit (with some safety margin applied)
3. If the endpoint is unavailable or the request fails, some constant default value may be put into model_memory_limit (e.g. 200MB)

Umbrella issue: elastic/elasticsearch#44699

[elasticmachine]

Pinging @elastic/ml-ui

[peteharverson]

For 7.4 the default model memory limit is set to 50MB and the setting is editable in the advanced editor.

For 7.5 we should look to auto-populate the model_memory_limit setting with the value obtained from the _ml/data_frame/analytics/_estimate_memory_usage API endpoint.

[alvarezmelissa87]

Closed via #50714