LQR by definition gives the optimal state-feedback law that minimizes certain quadratic objective function. In that sense, LQR is the best controller. However, LQR has a drawback that it assumes that all the states of the system are measurable. If not, an observer that estimates the states by observing the measureable output is required. with or without observer, LQR still requires an analytical model of the system. Additionally, if the system model is not linear, the design of LQR and observer mostly requires model linearization. So, with LQR, the design may be quite complex.
PID controller, in contrast, is simple and it can be tuned without having an analytical model. For research purposes, designing LQR for a complex system may result in a publication. On the other hand, for commercial applications, the PID controller is favourable for its simplicity and simple tuning procedures are already available.
the controller is better depends on the whether it achieve the desired design criteria or not, ofcourse LQR is better in a sense that it gives desired output with less control effort, but it requires lot of complexity to solve (for a starter) but PID controller is easy to build but it can be better that LQR if it combined the effect of LQR for e.g. See the paper "[14] He J.B., Wang Q.G., Lee T.H. PI/PID controller tuning via LQR approach. Chem Eng Sci 2000; 55(13):2429–39" .