distribution_based_score_function failing when n=1

[swatts-ci]

Hey @joein - realised this fix wasn't sufficient.

When we get len(response) == 1:

• We get a div by zero when calculating variance, as we are using Bessel's correction
• Even without Bessel's correction, we'd get a div by zero later on when scaling the scores

We should instead not use the response when len(response) < 2

Originally posted by @swatts-ci in #815 (comment)

[jan-kubica]

@swatts-ci resolved by #872

[joein]

The fix should've been available as of qdrant-client 1.13.0, closing as completed