Improve model standard detection from files

[sanjayankur31]

See: OpenSourceBrain/SBMLShowcase#77

At the moment, we detect files using their file extensions, but this isn't perfect because people may just use "xml" for files, in which case our tests will fail.

So, what we should do is:

• check the extension
• if that fails, for example, if it's "xml" which is ambigious, we read the first and last N lines of the file to grep them for tags