TCK streamlines gets stuck when file is not a TCK

[anibalsolon]

Currently, the nibabel.streamlines.tck.TckFile (haven't looked at the other formats) does not verify that a file is a .tck (i.e. has the "magic number" mrtrix tracks).

Since it does not verifies it, the code goes straight to try to read the key-value headers and get stuck in a loop, waiting for an "END". readline does not check for EOF.

If _read_header was a class method, it would be possible to do:

if magic_number.strip() != cls.MAGIC_NUMBER:
    msg = ("TCK magic number not found.")
    raise HeaderError(msg)

however, being a static method, one could only reference the TckFile.MAGIC_NUMBER (or any other way really).
Also, a better way to read the lines would be better, and detect EOF.

This is my new while true (and a test case for this issue):

echo "hello\nthere" > test.tck; python -c "import nibabel as nb; nb.streamlines.load('test.tck')"

[effigies]

Good catch. As the finder of the bug, would you like to propose the fix? We should definitely not be using readline to detect a fixed length magic number.