[Feature Request]: Secondary index about infinity

[yangzq50]

My proposal for the implementation:

1. When building secondary index for a numeric column in the table, we can build one secondary index structure for each segment of the table.

2. Every secondary index structure should contain 3 parts:
   2.1. pgm: a data structure which provides approximate range
   2.2. vector_value: sorted array of all the values of the chosen column in the segment (necessary for getting exact range)
   2.3. vector_offset: array of segment offsets for each value in vector_value (necessary for getting filtering results)

3. vector_value and vector_offset can be split into parts with max size DEFAULT_BLOCK_CAPACITY

4. Steps for range search of one column with secondary index:
   4.1. Get an approximate range from pgm
   4.2. Get the exact range by binary search in vector_value
   4.3. Generate a bitmask (when the size of the exact range is too big) or a vector of chosen offsets

[yingfeng]

4.2 is not necessary because pgm has already told the approximate range, as a result 4.2 is acturally a linear scan

4. Steps for range search of one column with secondary index:
   4.1. Get an approximate range from pgm
   4.2. Get the exact range by binary search in vector_value
   4.3. Generate a bitmask (when the size of the exact range is too big) or a vector of chosen offsets

[yangzq50]

subtask #476 and subtask #477 are deferred to #637