[optimization] Change SSTable serialized layout to enable varint and integer compression

[rajatkb]

The current SSTable layout works like the following : ref

	//  Then we have key-value pairs with length prefix
	//  ╭─────────┬────────────────┬────────────┬──────────────────╮
	//  │keyLength│ key            │ valueLength│ value            │
	//  ├─────────┼────────────────┼────────────┼──────────────────┤ ... repeat key-value pairs
	//  │4 bytes  │ keyLength bytes│ 4 bytes    │ valueLength bytes│
	//  ╰─────────┴────────────────┴────────────┴──────────────────╯
	//  Then we have 'index' which maps each key to its offset in the above key-value pairs
	//  SSTable.indexOffset refers to this point where 'index' begins

Suggestion to change the layout to the following

╭――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――╮
│key_length_byte_count (4byte)     │  <len 1 , len 2 .... >          │                                      
│value_length_byte_count (4byte)   │ <len 1 , len 2 .... >           │ 
│<key1 , key2 , key3 , key4 ....>  │ <value1 , value2 , value3 ...>  │
╰――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――╯

This would allow for

• Use a combination of varint + run length encoding on both the key and value length data which were static 4 bytes prior
• for both keys and values if they have a specific data type with certain kind of redundancy, e.g keys with prefixes or repeated null values we can employ some form of compression to handle redundant data.