feat(schema): annotated WDL schema, as JSON and YAML

[craigthackerx]

Adds schema/: the Azure Logic Apps workflow definition schema (2016-06-01) with annotations and corrections, published as both JSON and YAML.

Why

The published schema is machine generated: 147 KB, no definitions section, everything inlined through allOf/oneOf, and descriptions like "The flow triggers." that restate the property name. It validates a definition but does not teach one, and it says nothing about the behaviours that pass validation and then fail at run time.

It also rejects definitions Azure itself emits:

	Whole definitions validating cleanly
upstream schema	13 of 35
this schema	35 of 35

Measured against the 87 workflow definitions in the workspace, including this module's own examples, the azure-soc and az-socv5 SOAR estates.

What is in schema/

File	What it is
workflowdefinition.schema.json	upstream, committed so a rebuild needs no network
workflowdefinition.annotated.schema.json	the deliverable
workflowdefinition.annotated.schema.yaml	the same document in YAML, annotations as literal blocks
annotations.yaml	21 notes and 5 corrections, hand authored, the only file to edit
generate.py	fetches upstream, applies both layers, writes both outputs
README.md	how to point an editor at it, how to validate in CI, how to regenerate

The corrections

Each was found by validating real definitions, not by reading. Each widens what is accepted; none makes the schema stricter. Occurrences are how many real actions upstream rejected.

Correction	Occ.	What upstream gets wrong
retry-policy-type-casing	39	enum is None/Fixed/Exponential; designer and every example emit lowercase
authentication-managed-identity	28	the authentication union has no ManagedServiceIdentity branch, the recommended auth for an Http action calling Azure
retry-policy-count-expression	5	count typed integer, rejecting "@parameters('retry_count')"; any WDL value may be an expression string
variable-type-casing	3	InitializeVariable/SetVariable types PascalCase upstream, lowercase in practice
initialize-variable-multiple	1	variables capped at maxItems: 1; a portal export from a live workflow carries three

Each is recorded in x-annotation.corrections with pointer, reason and count, and called out in the description at the node it changed. They should be reported upstream and deleted when fixed.

Verification

• YAML output == JSON output (parsed and compared)
• the annotated schema passes Draft4Validator.check_schema
• 35/35 whole definitions validate, from both the JSON and the YAML file
• generate.py --check is a clean drift gate
• every annotation and correction pointer is asserted to resolve, so an upstream reshape fails the build

Note

The YAML is the schema in YAML, for readability and for tools that accept a YAML schema. A workflow definition itself is JSON: WDL has no YAML dialect, and the schema now says so where someone will read it.