Add support for parsing metadata from IO objects

[ikapelyukhin]

Changed the interface for all parser classes in order to support supplying IO objects (e.g., File or StringIO) as data sources.

All parser classes now don't take any constructor arguments, instead the parse and parse_file methods expect IO objects and filenames as arguments respectively.

Closes #15.

[ikapelyukhin]

@josegomezr give it a look, give it a test.

[felixsch]

Is this stale? Do you need a review @ikapelyukhin?

[ikapelyukhin]

Is this stale? Do you need a review @ikapelyukhin?

Hi, @felixsch. Yep, could use a review. What's the final word on this and StringIO, do you want to move ahead with this? I haven't heard back from @josegomezr, so I haven't merged yet.

[josegomezr]

@ikapelyukhin ah! I replied onm the Issue hehe, but looks great to me so far.

I left on the issue comment benchmarks and a sample impl of the parser working event-based: #15 (comment)

Benchmark-wise looks good, the event-based part could be handy in the future

[ikapelyukhin]

@ikapelyukhin ah! I replied onm the Issue hehe, but looks great to me so far.

I left on the issue comment benchmarks and a sample impl of the parser working event-based: #15 (comment)

Benchmark-wise looks good, the event-based part could be handy in the future

OK, let's merge this bad boy then.

Full blown event-based architecture is probably not required (would you really ever have more than one subscriber?), it's probably enough to just pass a code block to the parse method. I'll contemplate on that and maybe write something a bit later.