Use Random Generator in SMC instead of global seeding

[ricardoV94]

Closes #4938

Other changes:

1. Use model.recompute_initial_point instead of deprecated model.initial_point.
2. Update Metropolis Proposal classes to accept a Numpy Random Generator in __call__

[codecov]

Codecov Report

Merging #5131 (a80b650) into main (8a27ff4) will decrease coverage by 0.04%.
The diff coverage is 73.07%.

@@            Coverage Diff             @@
##             main    #5131      +/-   ##
==========================================
- Coverage   77.61%   77.57%   -0.05%     
==========================================
  Files          87       87              
  Lines       13981    13993      +12     
==========================================
+ Hits        10852    10855       +3     
- Misses       3129     3138       +9     

Impacted Files	Coverage Δ
pymc/smc/sample_smc.py	83.33% <37.50%> (-2.18%)	⬇️
pymc/step_methods/metropolis.py	82.90% <71.87%> (-0.80%)	⬇️
pymc/smc/smc.py	98.35% <100.00%> (-0.02%)	⬇️
pymc/tests/helpers.py	65.21% <100.00%> (+3.31%)	⬆️
pymc/backends/ndarray.py	70.12% <0.00%> (-0.19%)	⬇️
pymc/model.py	83.75% <0.00%> (-0.05%)	⬇️
pymc/distributions/distribution.py	94.53% <0.00%> (-0.03%)	⬇️
pymc/step_methods/mlda.py	96.38% <0.00%> (-0.02%)	⬇️

[aloctavodia]

LGTM!