rewrite node encoding for clarity and ease of extension

[missinglink]

this PR rewrites some older code in a clearer way, which opens it up for extension.
in particular issue #96 could benefit from this refactor

some notes on what this part of the code does:

• when 'node' elements from OSM are stored in leveldb they must be encoded/decoded from a slice of bytes []byte{}
• we use a custom encoding scheme to save disk space
• for the lat/lon values, we convert them to their IEEE 754 binary representation and truncate them to 6 bytes each, which is a nice tradeoff of precision vs. size (this consumes 12 bytes total)
• we optionally use a 13th byte to store metadata about entrances and accessibility
• there are still 4 bits reserved in that 13th byte for extension (such as storing information about the node role in a relation)

the subject of this PR is mainly to make the operations on the metadata byte more obvious and open them up for extension.

I didn't add new tests since https://github.com/pelias/pbf2json/blob/master/encoding_test.go seems to cover this sufficiently

Hey @paulmach if you get a second could you please do a quick review?