Support reduction of serologically typed GL String

Uses WMDA rel_dna_ser.txt for the corresponding version of IMGT database to produce serology mapping

Use sqlite3 database for reference data

Use sqlite3 database for data

Offload MAC codes from memory to sqlite3 database (natively supported by Python) to reduce
memory footprint. All MAC lookups happen through the db. The alleles and G group expansions
are still held in memory.

In addition, all generated data is saved as tables in the same database. This leads to one
file for storing all reference data in a standard format.

This led to drastic reduction in memory usage and startup time.

Version	First Time	Prebuilt Data
0.1.0	10.5 sec	4.92 sec
0.2.0	814 msec	598 msec
0.3.0	24.1 msec	24.7 msec

Heap memory used by ARD reference data after ard = pyard.ARD(3290)

Version	Memory (MB)
0.1.0	2977.86 MB
0.2.0	420.76 MB
0.3.0	3.74 MB

rearrange data in memory

This release rearranges how memory is used especially MAC codes and a lot of cleanup.

load_mac_file flag to load MAC file

Rename download_mac flag

• Rename download_mac ARD flag to load_mac_file as it properly describes what it does.
• Remove dead code
• Reformat code and fix some comments
• Version bumped to 0.1.0
• Updated pandas to 1.1.2

fix tests and sorting

fixes test and also a bug in 4th field sorting

allow P and G as input, fix lg and lgx behavior

This release fixes the behavior of lg and lgx to always reduce to 2-field.
It also allows P and G alleles as input

Fixes G-codes expansion

Fixes G-codes expansion when smart sorting.

Specify path for temporary files

You can specify path when creating ARD object.

ard = ARD('3290', data_dir='/tmp/py-ard')

version 0.0.16

update MAC location and version to 0.0.16

update to MAC location

0.0.15.0

update version (#37)