This is what comes to my mind... incomplete, but i hope it helps.
(1)
Do a "SIGNAL(1) - GND(2) - GND(3) - SIGNAL(4)" stack. Route Power on (1) and place only power-components/caps/inductors/drivers/connectors on (1). Place the MCU and its support (LDO, XTAL, whatever) on (4).
(2)
Do not split grounds. No benefit - only problems.
(3)
Do not route on (2) - if necessary, route on (3) but try to avoid it.
(4)
Spend 95% of your time on component placement - especially the power components.
Keep traces short and think about the current loop area.
(5)
If you connect GND-Pads on (1) to (2), use multiple vias and keep inductance in mind (DC-Resistance is an issue, but less important than via inductance).
(6)
Make sure to design the traces wide enough (use e.g Saturn PCB calculator). Think about adding PCB-solderable busbars (available via digikey etc...).
(7)
When you finished your layout, stitch the GND planes together with GND-vias. Be generous!
(8)
Think about test-pads/connectors: What measurements do you want to take? What test-points are required?
Is there a critical measurement? Maybe connect the signal via a BNC/SMA directly to your scope instead of using a probe?
(9)
Add some plastic stand-offs in the corner, so you can place the PCB on your desk nicely.
(10)
Think about safety: Yes, even in a lab. 100VDC is not a joke anymore, so maybe add standoffs on the top as well and place a plexiglass sheet above your PCB... Nice touch.
(11)
Think about the Silk-screen: Scratching your ehad in the lab before searching for the schematic PDF again takes time..... Use the silkscreen to your benefit.
(12)
Think about your connectors: How to connect your lab-supply/lab-sink/dut/pc nicely?
5mm banana? BNC? USB-A?