Estimating P-values from simulations

An employee at an aquarium monitors how much their sea otters eat. The amount of food a particular otter eats daily is approximately normally distributed with a mean of $17$‍ pounds and a standard deviation of $1$‍ pound. They suspected this otter was not eating enough, so they took a random sample $n=10$‍ days and observed a sample mean of $\overline{x}=16.5$‍ pounds of food per day.

To see how likely a sample like this was to occur by random chance alone, the employee performed a simulation. They simulated $40$‍ samples of $n=10$‍ values from a normal population with a mean of $17$‍ pounds and a standard deviation of $1$‍ pound. They recorded the mean of the values in each sample. Here are the sample means from their $40$‍ samples:

They want to test $H_0:\mu =17\text{ lbs}$‍ vs. $H_{\text{a}}:\mu <17\text{ lbs}$‍ where $\mu$‍ is the true mean amount of food per day.