Make `imu` class general to handle `p`-signals

[coatless]

As multivariate time series come into play, there needs to be a way to internally represents them. The best way forward is probably to generalize imu as it already has support for p signals. (albeit, with dependency requirements).

Linked to #192

Perhaps create a globally unifying class? Or maybe just cast as a ts object in R?

As of right now, we are stuck doing:

# Earthquake
data(EQ5, package = "astsa")
data(EXP6, package = "astsa")

EQ5.df = fortify(EQ5)
EQ5.df$type = "earthquake"
EXP6.df = fortify(EXP6)
EXP6.df$type = "explosion"

eq.df = rbind(EQ5.df, EXP6.df)

ggplot(data = eq.df, aes(Index, Data)) +
  geom_line() +
  facet_grid( type ~ .) +
  ylab("Displacement") +
  xlab("Time (seconds)")