Properly remove f.name usage in send_file

[untitaker]

Amend #1849

---

This change is

[untitaker]

There is a broken PR in master (#1849) that I've merged, I want somebody else to review this closely.

[untitaker]

FWIW, the original discussion is #104 -- I'm torn about mimetype support.

[mitsuhiko]

I'm okay removing etags I suppose but if we stop sending correct mimetypes that's not great.

[untitaker]

We don't have to remove anything but usage of f.name attribute I think (as far as I understand this is the only thing that is actually broken design)

[untitaker]

@mitsuhiko Check it out now, a.) Doesn't use f.name b.) Doesn't silently omit MIME-types if none can be guessed but just fails.

[mitsuhiko]

Looks good to me.

---

Reviewed 1 of 3 files at r1, 2 of 2 files at r2.
Review status: all files reviewed at latest revision, all discussions resolved, some commit checks failed.

---

Comments from Reviewable

[untitaker]

Fixed tests. I wonder if we should still infer the attachment_filename from f.name.

[untitaker]

Ah whatever.