Spec: Add non-REST API cookbook patterns

[a6b8]

Problem

FlowMCP v2.0.0 is designed for REST/JSON APIs, which covers ~70% of the APIs researched. However, many valuable data sources use non-REST protocols that could still be wrapped as FlowMCP schemas with appropriate handler patterns.

Non-REST Protocols Encountered

From researching 300+ APIs for FlowMCP schema issues:

Protocol	APIs	Handler Complexity
SDMX (Statistical Data)	ECB, Eurostat, IMF, BIS, Bundesbank, OECD	Medium — URL path encoding, XML→JSON transform
WFS/OGC (Geospatial)	Copernicus, EEA Waterbase, national geo services	Medium — XML/GML response parsing
SPARQL (Semantic Web)	Wikidata, EU Open Data, DBpedia, UNESCO	High — query language in request body
TAP/ADQL (Astronomy)	SIMBAD, VizieR, ESA archives	High — SQL-like query language
GraphQL	GitHub (some endpoints), Dune (new API)	Medium — query in POST body
SOAP/XML	Some government services, WSDL-based	High — XML envelope construction

Current Situation

The handler pipeline (preRequest → executeRequest → postRequest) is powerful enough to handle all of these, but schema authors have no guidance on:

• How to structure preRequest for query language injection (SPARQL, ADQL)
• How to transform XML/GML responses in postRequest
• Best practices for SDMX URL path construction
• When to use executeRequest override vs standard fetch

Proposed Addition

Add a cookbook section to the spec with handler patterns for the most common non-REST protocols:

1. SDMX Pattern — URL path encoding, XML→JSON, shared dimension lists
2. GraphQL Pattern — POST body with query, variables extraction
3. SPARQL Pattern — Query injection, result binding transformation
4. XML/WFS Pattern — Response parsing, GeoJSON conversion

Each pattern should include a complete minimal example schema.

Priority

P3 — Unlocks ~30% more API coverage. Not blocking for production but significantly expands the FlowMCP ecosystem reach.