This is not a highly pixelated image! The degree of granularity you are seeing is regular and ordinary for digital cameras imaging scenery at night. This granularity is called “noise” and it is caused by the camera’s software as it obligatorily apples amplification to the charges of photosites that otherwise would be void of detail due to under-exposure.
However, the chief failing you are seeing in this picture is what we call “flare”. Flare is caused by misdirected light intermingling with the image rays. The lens maker in his/her attempt to make a lens that yields a faithful image must resort to using multiple individual lenses in the lens array. The idea is to optimize the lens. Now each lens element of the array has two polished surfaces that act somewhat like a mirror. Each surface reflects away some light and this stray light bounces about. The net result is, some stray light baths sensor or film. This is the flare that reduces image contrast and generates star like rays that seem to spring from light sources and other highlights.
You shouldn't zoom in and focus. That lens is not parfocal, so when you zoom back out, the focus will shift slightly. Unless, of course, you mean that you're electronically zooming in on part of the picture, in which case, carry on. IMO, you're usually better off letting the camera handle focusing, because it will be more precise than you can practically be by hand.
A lot of other things, such as smog, can limit sharpness, though I have no idea if that happened here. Unfortunately, I really can't tell much from the photo that you posted because the JPEG artifacts are so severe that I can't tell how sharp it was originally. I'm assuming you shot with RAW. If not, always do that. If so, please update your question by uploading a cropped photo that includes a small part of the image at full resolution—preferably in PNG format so that it is lossless.
As Alan Marcus points out, there main issue here is noise, the best way to get rid of that is via image stacking. If you take 25 similarly exposed pictures, align them and then take the average then you reduce the noise by a factor of 5. Note that aligning is necessary even if all the pictures are taken on a tripod and remote control is used, after every shutter movement there will be tiny shift in camera orientation, enough to cause unsharpness at the pixel level). Here the noise reduction should be set to off, except the long exposure noise reduction which performs a dark frame subtraction.
Noise reduction eliminates not just the noise but all the invisible details that are hidden well below the noise floor. Taking the average of the 25 pictures will then not make these details emerge. Noise reduction should therefore be applied only after the averaging over the image stack is carried out.
The noise in each individual picture can be greatly reduced by exposing for longer, but that would overexpose the brighter areas. You can deal with this problem by taking a few different exposures and compile a HDR picture. You can e.g. take ten 30 seconds exposure at ISO 100 for the dark areas. The noise after averaging and adjusting the exposure here will be approximately ten times lower than in your picture (3 times longer exposure reduces the noise by a factor of 3, and averaging over ten images reduces the noise by a another factor of sqrt(10)). You can take ten exposures at 5 seconds exposure time, ten more at 1 second, ten more at 0.2 seconds etc. etc. until you get exposures that will correctly expose the brightest lights.
In total you may then need to take, say, 70 pictures, but this doesn't take a lot more effort than just taking a few picture. The end result will be a noise free HDR picture. While correctly exposed bright areas may not be the main goal, overexposed windows and lights in the picture cause so-called blooming effects where the charge in overexposed pixels leak to neighboring pixels. Dark areas directly adjacent to bright areas then become bright, the details there will get lost.
