And the last shall come first....
The problem with your plan is that there never has been any way for a GRAPHICS CARD to send to the mobo a GPU chip temperature, so that the MOBO can use that to control a fan aimed at the card. So that Noctua fan aimed that way can have its speed controlled a couple of ways.
(a) Connect it as you plan, to a CHA_FAN header, and configure that header to use PWM Mode of control and the MB (motherboard) temperature sensor. That is NOT the temperature inside the GPU chip by any means, but it does reflect the air temperature in the case.
(b) Connect the same, but set that header to use the CPU chip temperature sensor, betting the heat generation due to workload changes in the CPU chip will be VERY similar to heat generation and workload in the GPU chip.
(c) Forget control of this fan. Just set it to run full speed all the time, by powering it directly from the PSU with an adapter cable. This does NOT make any use of its PWM capability. It may over-supply air to the graphics card, but that does no harm.
(d) Connect as in (c) above, BUT look what came with the Noctua fan. SOME of their fans come with little items they call "LNA" for Low Noise Adapter. This is used when your fan is connected to a full 12 VDC power supply and has no speed control. It simply inserts into the fan power supply connector and is a resistor to reduce the fan speed (and airflow, and noise).
You have not told us details of your graphics card. Most actually have their own cooling fan built in, and most of those do their own fan speed control based on their GPU chip's internal temperature. They just cannot send those signals back out to the mobo for use there.
Your thoughts about powering the pump and fans of the Kraken X53 system are on the right path. The PUMP part is wired just like a 3-pin fan. Any 3-pin fan (and this pump) plugged into a 4-pin mobo header that is using PWM Mode for control (and your mobo's CPU_FAN header DOES do it that way) will always operate at full speed all the time. That IS what your pump is supposed to do. So I recommend that you get a simple 3-output 4-pin fan SPLITTER like this
https://www.amazon.com/Splitter-Computer-Extension-Converter-TeamProfitcom/dp/B07F8LV1BY/ref=sr_1_4?crid=1B5YJ1RA6HC0W&dchild=1&keywords=fan+splitter+4+pin&qid=1612756562&sprefix=fan+splitter,aps,183&sr=8-4
and plug that into your CPU_FAN header. Look closely at the three output arms. Only ONE of them has all four of its pins, and that is the only one that will send back to the CPU_FAN header the speed of its device. So plug into that one the 3-hole cable from your PUMP, then plug the two rad fans into the other outputs. The CPU_FAN header has an important second function: it monitors its device's speed signal for FAILURE. If that happens it will alert you immediately, and SOME mobos may take quick action to shut everything down even before the CPU internal temperature sensor shows high temps. In an AIO system, the critical component to monitor for failure is the pump. As long as it runs, even with partial rad fan failure, you get CPU cooling.
If you connect this way, the PUMP will run full speed all the time as designed and its signal will be monitored for failure. The RAD FANS both WILL be controlled for speed by the PWM Mode of the header, but that header will never "see" nor display those fans' speeds. From time to time YOU should verify they both still work.
So that leaves your two front 3-pin case fans to connect. If you are already using one CHA_FAN header for the Noctua aimed at the graphics card, connect them both to the other header using a Splitter. (Yes, you can use a 4-pin Splitter for this.) If not, you can connect each to its own separate header. In either case, since these are 3-pin fans, ensure that they are configured to use DC Mode and to use the MB temperature sensor.