Cache duplicate query_object calls

[hugovk]

Small performance improvement.

query_object is called once per release file:

    for rfile, file_desc, os_slug, add_download, add_desc in release_files:
        if not os_slug:
            continue
        os_pk = query_object(args.base_url, "os", slug=os_slug)

There might be 12 release files but four slugs:

	File	Slug
1	Python-3.14.5.tgz	source
2	Python-3.14.5.tar.xz	source
3	windows-3.14.5.json	windows
4	python-3.14.5-embed-amd64.zip	windows
5	python-3.14.5-embed-arm64.zip	windows
6	python-3.14.5-embed-win32.zip	windows
7	python-3.14.5-amd64.exe	windows
8	python-3.14.5-arm64.exe	windows
9	python-3.14.5.exe	windows
10	python-3.14.5-macos11.pkg	macos
11	python-3.14.5-aarch64-linux-android.tar.gz	android
12	python-3.14.5-x86_64-linux-android.tar.gz	android

So we only need to call the API four times, not 12.