So, cooling first.
Actually, a GOOD air cooler with high quality, low noise fans, like those sold by Noctua or Thermalright (Not to be confused with Thermaltake), are generally QUIETER than liquid cooling. For one thing, the fans tend to run at lower RPM than on most AIO coolers, but of course whether using air or water you can make some adjustments to the fan profiles to tailor this to your personal preferences somewhat, but what you really CAN'T do a lot about is the sound of the pump on an AIO or custom loop. When you combine fan noise with pump noise, you are almost always looking at a moderately louder configuration with most AIO coolers.
With a very good air cooler, you should almost never even hear it except when you are under a pretty demanding load, especially if you factor in the already present sound of any case fans. My system which has five Noctua 140mm case fans and a Noctua NH-U14S with two more of those same NF-A14 PWM chromax.black.swap Noctua fans on it, is basically silent unless I run a VERY demanding game or a stress test like Prime95. The rest of the time, from about three feet away, it is absolutely silent.
Most noisy systems are due to either the presence of low quality fans that are not well engineered, or use poor quality bearings, OR are simply not configured with a reasonable fan curve in the BIOS or whatever is used for fan control on that system, because most people just use the presets and even the silent preset on most systems is garbage that results in a noisy system. If you have high quality fans I can show anybody how to set up their fan curve so that 90% of the time it is silent and even when under a substantial load isn't unbearably noisy unless there is a thermal issue or the cooling configuration is not adequate for the hardware.
There is basically NO difference in performance, yet, between DDR4 and DDR5 unless you have a system that can handle the very top end memory kit speeds. At some point that may change once they get the latency down, because if you compare, say, 3200MT/s DDR4 with a CL14 latency to a 5600MT/s DDR5 kit with a CL40 latency, the latency pretty much kills the benefit of the higher frequency. This has always been true for memory until a new architecture has matured enough that memory manufacturers are able to get the CL latency down to a level where those higher frequencies actually start to make a difference in the resulting True Latency or "First word" latency. I could show you how that works, but it goes well beyond what you were asking so I won't unless you really want to see the formula for determining that. Plus, DDR5 is significantly more expensive.
If you run a DDR4 system with 3200-3600MT/s frequency and a low latency, the chances are very good you will have similar true latency, which is what actually determines how fast memory operations are and how "snappy" the system feels in regard to memory performance, as you will to most of the supported DDR5 configurations out there. That IS beginning to change, but you would want a rather high end board, higher end CPU and very good memory kit to even make it worth it and even then the difference for most use cases will not be that significant. The more important aspect of it is probably only whether you want to go with a DDR5 system for future proofing, to some degree, since DDR6 is already on the horizon anyhow, since there are no more consumer platforms going forward that will have DDR4 support after the current Intel generation has run it's course. All future Intel and AMD platforms will be using DDR5 or higher.
But if this system lasts you five years, there will be something ELSE that is mainstream by then anyhow, so really it doesn't matter unless you plan to again at that time buy a platform that will still be using DDR5 which will likely mean not whatever is actually the current Gen hardware at that time.