Have not designed something like this.
But trick would be get the floating current source right wayLast string is the ouput stage as we all can see, The once before it are for biasing of the floating current source in the output of folded cascode.
you can increase the openloop gain of the opa, which can increase the thd, please take care of the impedence of float current source, the impedence will increase as the frequence increase. you can add a capacitor between the float current source.