Estimate best kernel parameters resetting `curr_points`

[Balaji Pokuri]

Observed during bo_md speedup calculation. When the kernel parameter estimation fails, the value of `curr_points` is being reset to n_init . So, PARyOpt will start failing if the kernel parameters are estimated frequently.

‌

Other functionality is working without any issue.

[Balaji Pokuri]

Issue is in _is_near_to_population, where nearness is being calculated as a function of theta and not absolute distance. Can cause issues if theta is large, when new points are no longer added and hence loss of data.