Worth warning about memoising random functions

[MCMaurer]

Hey there,

This isn't necessarily a problem with memoise at all, but rather a warning that might be worth giving to folks who read about the package.

If someone tries to memoise a function that includes any sort of randomization in it, their outputs will not be random if the inputs are identical, and it could be really hard to tell just by looking at the results. Here's a simple example:

rnorm_mem <- memoise(rnorm)
x <- rnorm_mem(10)
y <- rnorm_mem(10)
all.equal(x,y)
#> [1] TRUE

It seems pretty obvious, but I could definitely see someone getting a little trigger-happy with memoise and making this mistake without realizing it.

[katrinleinweber]

Does this also occur, if the memoised function receives inputs in a random order? Would explain #106.