[CT-2752] Run DSI SemanticManifest validations during parsing

[QMalcolm]

In dbt-semantic-interfaces there lives a tool called SemanticManifestValidator. The SemanticManifestValidator takes in a semantic manifest and runs a series of rules against it to ensure that it is valid. These are things like ensuring that either a measure has an agg_time_dimension set or that there is a default_agg_time_dimension set on the semantic_model that the measure belongs to.

Currently we don't run these validation rules in core. The only way someone using MetricFlow with core can run the validations is by doing

$ dbt parse
...
$ mf validate-configs

The issue with this is that validations are only run after the semantic manifest is written. This is problematic because validation issues should cause a failure before writing the semantic manifest. Additionally most people likely won't run mf validate-configs separately. Which means the issues would only be caught at query time 😬 We want to be informing customers of their issues when writing their configs, thus we need to ensure the SemanticManifestValidator is run during parsing.

Luckily, using the SemanticManifestValidator is fairly easy. It's a generic python class that requires that you declare an implementation of the SemanticManifest protocol on instantiation. It would look something like

validator = SemanticManifestValidator[SemanticManifestImplementation]()
validation_results = validator.validate(semantic_manifest)

The other part of this is presently we only construct the SemanticManifest, and specifically a PydanticSemanticManifest right before we write out the semantic manifest. We can either reuse the PydanticSemanticManifest builder to get a PydanticSemanticManifest and run that through a SemanticManifestValidator or we'll need to actually create a core implementation of the SemanticManifest protocol and then run that through a SemanticManifestValidator. The former is easier, but the latter is probably more in core's style.

Acceptance Criteria

1. If a manifest has semantic models and/or metrics, the SemanticManifestValidator should get run as part of parsing
   • It needs happen after nodes have been fully "processed" (i.e. after things like this)
   • It needs to happen before writing out the semantic manifest
2. Validation issues returned by the validation process should get logged
3. If there are ERROR level validation issues, parsing should be considered as having failed.