Implement lighter-weight archive format

[JonathonReinhart]

I've wasted so much time hacking on/around libtar, that it would have made more sense to implement our own lightweight format.

• Python writer (and probably reader for debugging)
• C reader for extraction

The format would be focused on a fast, simple C extractor.

I'm thinking something like this:

File header:

Offset	Size	Type	Name	Description
0	4	uint32	magic	Sanity check

(We don't need a version or flags, since the archive is attached to the extractor)

Entry:

Offset	Size	Type	Name	Description
0	1	uint8	ent_type	Entry type (1 = file, 2 = symlink)
1	1	uint8	comp_type	Data compression type (0 = none, 1 = lzma )
2	2	uint16	name_len	Length of filename
4	4	uint32	data_len	Length of data
8	name_len	utf8_string	name	Filename / path
8 + name_len	data_len	raw	data	Raw data (may be compressed) Filie data or symlink target path

We don't need uid/gid/timestamps or any of that crap. The only thing we may need is executable, but we could just as easily set everything executable.