Compression Support & New sigmf.tofile method

[Teque5]

Closes #26. Adds read/write support for compressed SigMF archives (.sigmf.gz, .sigmf.xz, .sigmf.zip) using only stdlib so no new dependencies are needed. Compression uses tarfile for gzip/xz and zipfile for zip.

Also fixed performance issue where reading uncompressed .sigmf archives was extracting data to RAM — these are now numpy memmapped directly into the tar.

Also added a sigmf.tofile() convenience function to mirror sigmf.fromfile(). It infers the SigMF datatype from the numpy dtype and handles compression and archive creation via kwargs:

data = np.array([<SOME_STUFF>], dtype=<SOME_DTYPE>)
sigmf.tofile("recording", data, sample_rate=48000)                       # loose .sigmf-data + .sigmf-meta
sigmf.tofile("recording", data, sample_rate=48000, toarchive=True)       # .sigmf archive
sigmf.tofile("recording.sigmf.xz", data, sample_rate=48000)              # compressed archive