gp_minimize: ranges can be a different datatype than initial guess

[ChrisBNEU]

I am using skopt version 0.9.0 and numpy 1.23.5 (due to an error I got that is specified in #1138). I found, after downgrading numpy from 1.24.1, that I had made an error in specifying the ranges. They were ints instead of floats. however, my initial x-guess was a float. I made a minimal example below:

from skopt import gp_minimize
def find_sqrt(x):
    """find square root of value y given a guess x"""
    y = 12.5
    x = x[0]
    y_guess = x**2
    return abs(y_guess - y)

x0 = [1.0]
y0 = None
    
res1 = gp_minimize(
    find_sqrt,
    [
        (0.0, 12.0),
    ],
    n_calls=12,
    x0=x0,
    y0=y0,
)

print("guesses with proper float ranges: ", res1.x_iters)

res2 = gp_minimize(
    find_sqrt,
    [
        (0, 12),
    ],
    n_calls=12,
    x0=x0,
    y0=y0,
)

print("guesses with improper int ranges: ", res2.x_iters)

which outputs:

guesses with proper float ranges:  [[1.0], [8.114281197171252], [9.021416917982528], [4.955995220826303], [6.649848371121621], [11.582469882107137], [3.953831466277097], [2.1390536274803877], [2.6277353173592157], [5.884289789665528], [3.918451377145078], [3.5048487566802144]]
guesses with improper int ranges [[1.0], [6], [9], [5], [0], [11], [1], [1], [7], [3], [4], [4]]

If this is expected behavior and I missed something in the docs then I am fine with closing the issue, but I would think that there would be an error thrown if the first and second guesses are different datatypes for the same variable.