besselj_zero(51, 1) is correct, but besselj_zero(52, 1) is not

[jlapeyre]

When nu is large enough, the lower order zeros are incorrect.
besselj_zero(n, 1) is correct for (integral) nu<52. But,

julia> besselj_zero(51, 1)
58.16033705734714  # correct

julia> besselj_zero(52, 1)
22.886532747688847  # incorrect

julia> besselj_zero(52, 2)
64.9539962041836   # correct

julia> besselj_zero(66, 2)
79.89174526702548  # correct

julia> besselj_zero(67, 2)
74.79148111822263 # incorrect

This might be fixed by using a bracketed root solver.

[jlapeyre]

We changed from fzero to find_zero in #17.
The bug in this issue comment is no longer present.