[Lut HAL] documentation update.

     * For data precision, 32-bit float is used to specify a Lut by both the HWC and

     * the platform.

     *

     * For unflattening/flattening 3D Lut(s), the algorithm below should be observed

     * by both the HWC and the platform.

     * Assuming that we have a 3D array `ORIGINAL[WIDTH, HEIGHT, DEPTH]`, we would turn it into

     * `FLAT[WIDTH * HEIGHT * DEPTH]` by

     *

     * `FLAT[z + DEPTH * (y + HEIGHT * x)] = ORIGINAL[x, y, z]`

     *

     * Noted that 1D Lut(s) should be gain curve ones

     * and 3D Lut(s) should be pure color lookup ones.

     * Note that 1D Lut(s) should be gain curve ones and 3D Lut(s) should be pure color lookup

     * ones. For 3D Luts buffer,the values of the lut buffer should be normalized, ranging from 0.0

     * to 1.0, inclusively and the data is organized in the order of R, G, B channels.

     * For 1D Luts, the lut's values should be also normalized for fixed point pixel formats,

     * and we now ignore floating point pixel formats + extended range buffers.

     */

    @nullable ParcelFileDescriptor pfd;