Add optional sequential mode for efficiency

[John-194]

Can I add an option to set sequential mode?
Sequential can be more efficient because it does not have the scheduler's overhead, but it depends on the dataset. Here are the results from my testing:

PARLAY_NUM_THREADS: Not set
  ordered point cloud:
     New threaded:   1.23 ms; CPU: ~650%;
     New sequential: 1.23 ms; CPU:  100%; efficiency: 6.5x
     Original:       1.23 ms; CPU: ~650%; 
  shuffled point cloud:
     New threaded:   1.46 ms; CPU: ~650%
     New sequential: 2.31 ms; CPU:  100%; efficiency: 4.1x
     Original:       1.45 ms; CPU: ~650%
     
PARLAY_NUM_THREADS: 1
  ordered point cloud:
     New threaded:   2.64 ms; CPU:  100%
     New sequential: 1.23 ms; CPU:  100%; efficiency: 2.1x
     Original:       2.64 ms; CPU:  100%
  shuffled point cloud:
     New threaded:   3.70 ms; CPU:  100%
     New sequential: 2.33 ms; CPU:  100%; efficiency: 1.6x
     Original:       3.69 ms; CPU:  100%

Using sequential mode is faster than setting PARLAY_NUM_THREADS to 1.
To make this work, I used a simple if check in the main loop, but the performance cost is within a margin of error - unnoticeable.

Usage:

import dbscan

# set sequential, disable threading
dbscan.set_sequential()
print("sequential:", dbscan.get_sequential())
dbscan.DBSCAN(points, eps=1.5, min_samples=15)

# enable threading again
dbscan.set_sequential(False) 
print("sequential:", dbscan.get_sequential())
dbscan.DBSCAN(points, eps=1.5, min_samples=15)