fix: add .mjs to AST extractor dispatch table

[shaun0927]

Partial fix for #387 (and #365).

.mjs was added to CODE_EXTENSIONS in detect.py (v0.4.15) but not to the _DISPATCH table in extract(), so .mjs files are detected and counted in corpus metrics but silently produce zero nodes/edges during extraction.

Change

One line: ".mjs": extract_js added to _DISPATCH in graphify/extract.py.

.mjs is standard ES module JavaScript — same language as .js, just with explicit module syntax. tree-sitter-javascript handles both identically.

Testing

from pathlib import Path
from graphify.extract import extract
import tempfile

tmpdir = Path(tempfile.mkdtemp())
mjs = tmpdir / "app.mjs"
mjs.write_text("""
import { readFile } from "fs/promises";
export function parseConfig(path) {
    return readFile(path, "utf-8").then(JSON.parse);
}
export class AppServer {
    constructor(port) { this.port = port; }
    start() { console.log("started"); }
}
""")
result = extract([mjs], cache_root=tmpdir)
print(f"Nodes: {len(result['nodes'])}")  # 5 (was 0 before fix)
print(f"Edges: {len(result['edges'])}")  # 5 (was 0 before fix)

Full test suite: 433 passed, 0 failures.

Note on .ejs

.ejs has the same dispatch gap but EJS is an HTML templating engine, not pure JavaScript. It may be better suited to DOC_EXTENSIONS for semantic extraction. Left for separate discussion in #387.