openapi: encoding allowReserved and extensions are dropped silently

[OmarAlJarrah]

Summary

An Encoding Object declares six fields. contentType, headers, style and explode all reach
ir.PartEncoding; allowReserved and x-* reach nothing at all — no IR field, no Unmodeled
entry, no diagnostic. A document declaring either compiles to IR identical to one that does not.

This is the same mechanism as the parameter-level allowEmptyValue gap: a declared serialization
keyword whose neighbours have IR homes while it has none, and which nothing reads. ir.PartEncoding
has ContentTypes, Headers, Multi, Filename, Style and Explode — no AllowReserved.

encodingConfig in compilers/openapi/internal/operation/content.go is the whole reader:

pe.ContentTypes = splitContentTypes(enc.GetContentTypeValue())
headers, diags := lowerHeaders(c, ts, anchors, enc.GetHeaders(), encPtr)
pe.Headers = headers
if enc.Style != nil {
    pe.Style = string(*enc.Style)
}
pe.Explode = enc.Explode

enc.GetAllowReserved() and enc.GetExtensions() are never called.

Reproduction

openapi: 3.1.0
info: {title: T, version: "1"}
paths:
  /x:
    post:
      operationId: postX
      requestBody:
        required: true
        content:
          multipart/form-data:
            schema: {type: object, properties: {f: {type: string}}}
            encoding:
              f: {contentType: text/plain, allowReserved: true, x-vendor: v}
      responses: {"200": {description: ok}}

Compiles with no diagnostics at all. The resulting PartEncoding is
{"contentTypes":["text/plain"],"multi":false,"filename":false} and Content.Unmodeled is empty —
both declarations gone without trace.

Expected

Each either lowers into an IR field or is kept verbatim under Unmodeled with ReasonNoIRHome and
a degraded-construct diagnostic, as allowEmptyValue now is at the parameter position. Note that
x-* is legal on every OpenAPI object and ir-design.md says it maps to namespaced Unmodeled,
so the extensions half may be better handled with the response-extensions work in #275 — the
allowReserved half is distinct from that.

Found while reviewing #272, which closes #39 and covers the parameter position of this mechanism
but not the encoding one.