[Bug] Incorrect inference of image type from URL

[leonceaklin]

Describe the bug
When generating Atom and JSON feeds with the feed library, the image MIME type is incorrectly inferred from the image URL even when an explicit type is provided. This results in malformed type strings such as image//328314f6-4317-4579-990c-b2047d1804ef/public when the URL includes segments like /public. RSS output appears to respect the provided type and does not exhibit this issue.

To Reproduce

1. Create a Feed instance with an image and/or add an item with an image, supplying an explicit MIME type (e.g., image/jpeg).
2. Generate Atom (atom1()) and JSON (json1()) outputs.
3. Observe that the image type in Atom/JSON is mangled based on URL segments, e.g., image//328314f6-4317-4579-990c-b2047d1804ef/public, while RSS (rss2()) behaves correctly.

Minimal example:

import { Feed } from "feed";

const imageUrl = "https://example.com/328314f6-4317-4579-990c-b2047d1804ef/public";

const feed = new Feed({
  title: "Test Feed",
  id: "https://example.com/",
  link: "https://example.com/",
  image: imageUrl,
});

// Add an item with an explicit image type (expected to be respected across formats)
feed.addItem({
  title: "Post 1",
  id: "https://example.com/posts/1",
  link: "https://example.com/posts/1",
  image: imageUrl,
  // Depending on API surface, explicit MIME type may be attached via attachments or extensions, e.g.:
  // attachments: [{ url: imageUrl, mimeType: "image/jpeg" }]
});

const rss = feed.rss2();    // OK
const atom = feed.atom1();  // BUG: type mangled, e.g., "image//328314f6-4317-4579-990c-b2047d1804ef/public"
const json = feed.json1();  // BUG: type mangled

console.log({ rss, atom, json });

Expected behavior

• If an explicit image MIME type is provided (e.g., image/jpeg, image/png), use it as-is across RSS, Atom, and JSON.
• Do not infer or transform the type from the URL when an explicit type is supplied.
• Ensure output contains a valid MIME type and does not include URL path segments.

Actual behavior

• Atom and JSON outputs construct an invalid type based on URL segments, resulting in strings like image//328314f6-4317-4579-990c-b2047d1804ef/public.
• RSS output respects the provided type.

Additional context

• The issue appears in Atom/JSON serialization paths but not in RSS, suggesting divergent type handling.
• It seems tied to URL-based inference (possibly a regex/path parsing bug) that concatenates URL segments into the MIME type.
• Proposed fix:
  • If a type is provided, skip inference and use the explicit value.
  • If no type is provided, infer from file extension or content-type heuristics, but never mix URL segments into the MIME type.