export: cache serialized values

[mm-ipinfo]

Make use of the new v2 API (#47) to cache decoded and serialized MMDB values.

Testing on a 1GB MMDB file:

┌────────┬───────┬───────┬─────────┐
│ Format │  Old  │  New  │ Speedup │
├────────┼───────┼───────┼─────────┤
│ CSV    │ 1m31s │ 27.2s │ 3.3x    │
├────────┼───────┼───────┼─────────┤
│ TSV    │ 2m18s │ 26.1s │ 5.3x    │
├────────┼───────┼───────┼─────────┤
│ JSON   │ 2m51s │ 15.1s │ 11.3x   │
└────────┴───────┴───────┴─────────┘

@max-ipinfo this is the raw Claude output which I might cleanup a bit, but it seems to work as expected.

[mm-ipinfo]

@max-ipinfo extracted CSV/TSV/JSON exports to dedicated exporter_* files:

type exporter interface {
	WriteRecord(result maxminddb.Result) error
	Flush() error
}

Also simplified a little the JSON logic. It's a bit slower (15s -> 24s) but still much faster than the current version.

[max-ipinfo]

It's a bit slower (15s -> 24s) but still much faster than the current version.

Just curious. Why would moving it to an exporter interface make it that slower? Are the function calls via the interface that much worse?

[mm-ipinfo]

Oh sorry, I was unclear 😄 It's not the interface but the JSON simplification that slows down things a bit. Instead of manually writing {"range": …, and then write the cached JSON (minus the first char) we just cache everything but with range set to __RANGE__ and we replace it with the actual value at write time. It's a bit slower due to the bytes.Replace call but it's also much less hacky.