I am not quite sure why you think your code should give you the results you are looking for.
The simplest approach would be to note that the median of the normal distribution is equal to the mean, so you could use your 50% quantile as an estimate $\hat{\mu}$. Next, for a normal distribution the 10% quantile is 1.28 times the standard deviation away from the mean, so you could estimate
$$ \hat{\sigma} := \frac{\hat{\mu}-q_{0.10}}{1.28}. $$
For slightly higher stability, you could estimate $\sigma$ in the same way using the 90% quantile, and take the average of the two estimates.
Note that this is very crude, and leverages only a few of your data points. But it has the advantage of being extremely simple to understand.
The "right" way to do this would be to numerically optimize both estimates, with an objective function that minimizes the difference between the quantiles of the fitted distribution and the quantiles you feed in. This will be more accurate. It will also be more complicated, more error-prone, and much slower.