feat: bodyweight data

[m-muecke]

Closes: #7

@fabian-s note the following:

• I've removed the ex (exercise) column since I've filtered for the bicycle crunch data. Perhaps, naming the dataset bicycle might make more sense.
• I've converted the factor variables to factor().
• I've not included the raw zip files, since the files are on GitHub, we could add a download script incl. unzipping in the beginning to make it more reproducable.
• we could also bucket the ratings like @HerrMo did, basically the following:

df |> mutate(coarse_rat = case_when(
  rating <= 1.5 ~ 1,
  rating > 1.5 & rating <= 2.5 ~ 2,
  rating > 2.5 & rating <= 3.5 ~ 3,
  rating > 3.5 & rating <= 4.5 ~ 4,
  rating > 4.5 ~ 5
))

[fabian-s]

thanks max! i think it might make more sense to add these data once we have support for multivariate functions in tf so that we can save these as x-y-z accelerations / rotations (instead of 6 different tfd's..)? not sure...

in any case I've extended this for the different exercises and tried to make the code a little more robust, see 4b7b79f

also not sure about the data quality here -- e.g. for squats looking at the first set of the first subject

filter(kn, id == "1" & set == "set1") |> ggplot() + geom_spaghetti(aes(y = acc, col = ordered(rep))) + facet_grid(loc ~ dim)

-- how can the y-acceleration of the left leg be always positive and the y-acceleration of the right leg always be positive (and constant)? that's a weird ass squat! should go down, then up, fairly synchronized between the two legs, no?