GPU pricing endpoint improvements

[Redm4x]

• Fixed the median calculation when there is even number of providers (use average of the two middle values)
• Add weighted average based on provider's gpu count
• Return price object of null if there is an issue during price calculation
• Cap allocated gpu amounts to allocatable amount to prevent numbers going into the negative when feature discovery return wrong values
  • provider reports excessively high amount of Allocatable cpu & ram when inventory operator hits an ERROR support#192