Given a stream of data, it is possible to determine the beginning of
EXIF data but not the end. Therefore, either an image-aware
implementation must know how to parse an image and extract the EXIF
data or a brute-force search implementation (one of which is provided
by this project) must find the start anchor and then return all bytes
from that to the end of the file.
We have been made aware of some use-cases where a brute-force search
might be unavoidable due to trust or stability issues with the image
structure. This leads to large allocations. This can be avoided by
accomodating support that will allow for both a byte-slice or an
`io.ReadSeeker`. Since the EXIF structure is typically not read-
intensive (a couple of kilobytes if no thumbnail is present), this
should have a minimal performance impact.
Closes#42