Fix integer underflow in try_range_response for empty files

[Darksonn]

Thanks to @mufeedvh for reporting this bug.

[seanmonstar]

Just to add a note from spec:

For a GET request, a valid bytes range-spec is satisfiable if it is either:

• an int-range with a first-pos that is less than the current length of the selected representation

So, with a file of 0 bytes, the first-pos of 0 is not less than the length, and thus is now rejected.

[yanns]

From CI:

error: the borrowed expression implements the required traits
   --> axum-extra/src/response/file_stream.rs:615:34
    |
615 |         std::fs::write(filename, &[]).unwrap();
    |                                  ^^^ help: change this to: `[]`

[Darksonn]

Fixed thanks.

[darth-raijin]

Small question, feel free to ignore it was meant as food for thought

[darth-raijin]

Question:
I noticed that the error body "Range Not Satisfiable" is hard-coded directly into the response. Would it make sense to extract this into a shared constant so tests can assert on the exact same value?

Having the literal inline makes tests fragile, since they need to duplicate the exact string. If the message ever changes, that could also unintentionally break consumers relying on the current wording or cause test drift.

In particular, there are two branches in this PR that both return StatusCode::RANGE_NOT_SATISFIABLE. It might be useful if the test also asserted on the error body, to ensure the StatusCode::RANGE_NOT_SATISFIABLE comes from the branch in file_stream.rs with the expected response body, and not the test branch.