Performant Python Patterns Notebook changes

[Robadob]

Whilst writing the PPP talk many of the examples were refined for the accompanying notebook. These should be reviewed to update those within the course.

Also opted to ignore the binary search example, it's difficult to justify a use-case.
Included a very brief @numba.njit example, this could be further developed.

[Robadob]

Following my talk two people made suggestions:

• Paul Graham (NVidia) said they have alot of resources about super easy python-gpu accelerate stuff, like drop in cudf replacement for pandas. Directed him to leave more details here, can follow up if he fails to do so.
• Someone random suggested that generator functions could be more performant than for-loops, I have a feeling I previously explored this and they're only recommended for saving memory, particularly with open ended streams of patterns.

[Robadob]

Giving PPP notebook a read through, looking at how it's examples can be better standardised

• Built in fns
  • Best/avg/worst case
  • Standardise this closer to other benchmarking, such as lists
• Uniques
  • Is there a better way to estimate duplicates than range vs input range
• Tuples
  • Timeit allows you to pass code as a string, this could therefore be automated to run through a wider range of list/tuple len.
• Misc
  • Some common utility graph function for graphing the scaling