Improve consistency of notation of scale in the examples

[HKaras]

Currently in the examples for parametric fitting there is a parmater called scale which is the integral of the distribution. When a non-parametric distribution is used this is called Pscale.

For example in the bimodal Gaussian example

Pfit = results.evaluate(Pmodel,r)
scale = np.trapz(Pfit,r)
Puncert = results.propagate(Pmodel,r,lb=np.zeros_like(r))
Pfit = Pfit/scale
Pci95 = Puncert.ci(95)/scale
Pci50 = Puncert.ci(50)/scale

# Extract the unmodulated contribution
Bfcn = lambda mod,conc,reftime: scale*(1-mod)*dl.bg_hom3d(t-reftime,conc,mod)
Bfit = results.evaluate(Bfcn)
Bci = results.propagate(Bfcn).ci(95)

but in the basic 5p DEER example:

# Extract fitted distance distribution
Pfit = results.P
Pci95 = results.PUncert.ci(95)
Pci50 = results.PUncert.ci(50)
Pfit =  Pfit

# Extract the unmodulated contribution
Bfcn = lambda lam1,lam5,reftime1,reftime5,conc: results.P_scale*(1-lam1-lam5)*dl.bg_hom3d(t-reftime1,conc,lam1)*dl.bg_hom3d(t-reftime5,conc,lam5)
Bfit = results.evaluate(Bfcn)
Bci = results.propagate(Bfcn).ci(95)

I propose chaing scale to Pscale so that code between examples can be more easily compared.