Benchmark/Bleachermark design

Having a Bleachermark with Benchmarks with semantically different pipelines doesn't make sense
So disallow different lengths and different function labels
So remove labels from Benchmark and put them on Bleachermark
User message: Benchmark is mostly an internal thing. Bleachermark is the core external class.

Logic for running benchmarks

Current proposal: Runners

Bleachermark as iterator

Should a Bleachermark have iterator behaviour?
If so, should we store the data when a Bleachermark is used as an iterator?

Parallelism

Suggestions?
Take a look at https://wiki.python.org/moin/ParallelProcessing.
Decouple the SageMath parallel framework into a separate package
It should be possible to make a speed plot across different no. of workers

Interruption and Resume

Works?
How to integrate with parallelism?
What did David do in #20684

Out-of-the-Box Statistics

Remove plural 's' from statistics names?
Rename stdvs to std?
Getters for a specific benchmark? (useful for plotting)

Other

Write setup.py and so on to make this a pip installable package
Decide on memory safety for Bleachermark.add (how do we avoid copying data on add'ing)?

User stories

Coding theory decoding setting
- Plot decoding speeds (x: noise/no. errors, y: speed)
- Plot decoding success (x: noise/no. errors, y: #correct/#trials)
Cost of sorting algorithms
- Plot sorting speed of two sorting algorithms (x: log of list size, y: log of time)
- Guess asymptotic complexity
- Adaptively determine size of trials
Numerics?