Archive index rework to make loading faster

[macneale4]

The initial impl of archive indexes over optimized for space. This resulted in being 10x slower to load the index of archives than noms table files. To address this:

• Dropped the end to end compression of the index
• Dropped the use of var ints for offset deltas and chunk refs
• Altered the use of byte span offsets, and instead used a end-offset approach which requires no delta processing on load.
• Used only slices of primitive types in the index memory. Constant time read path with a little more complexity, but allows us to read directly off disk into memory.

Testing indicates that on a 41 Gb archive file, this returned load performance to match classic table files, and the size of the index increased by about 350Mb (total ~ 1Gb)

[coffeegoddd]

@macneale4 DOLT

comparing_percentages
100.000000 to 100.000000

version	result	total
5ef8ca5	ok	5937457

version	total_tests
5ef8ca5	5937457

correctness_percentage
100.0

[coffeegoddd]

@coffeegoddd DOLT

comparing_percentages
100.000000 to 100.000000

version	result	total
0b83495	ok	5937457

version	total_tests
0b83495	5937457

correctness_percentage
100.0

[coffeegoddd]

@macneale4 DOLT

comparing_percentages
100.000000 to 100.000000

version	result	total
24185cc	ok	5937457

version	total_tests
24185cc	5937457

correctness_percentage
100.0

[github-actions[bot]]

Additional work is required for integration with DoltgreSQL.

[coffeegoddd]

@macneale4 DOLT

comparing_percentages
100.000000 to 100.000000

version	result	total
b4f995a	ok	5937457

version	total_tests
b4f995a	5937457

correctness_percentage
100.0