As
@Phipli already stated, that's not a practical possibility, because you'd need to create an 040 bus to PCI bridge, which would be an absurdly expensive thing to do, to say nothing of the engineering costs associated with bringing something like that into existence.
I agree that it's (emphatically) not a practical possibility, but the culprit would be the software IMHO (as it almost always is). Xilinx has both PCI and PCIe root port available for their FPGA, and there is bridge chip available both ways. Hardware-wise, mapping a PCIe root port in the slot space of the Mac and using the superslot for the PCIe space would probably not be particularly complex. Adding a SATA PCIe chip would be easy. Artix-7 have native pins for PCIe (GTP), so any board breaking those out would be enough. The Trenz TE0712 is would be my choice (I've looked at it for a hypothetical SBusFPGA 2.0). You could probably assemble a board like that for less than 500€, with most of the cost sunk into the TE0712.
However, for SATA on 68040 (or even '030), a shortcut would be to use said GTP pins to drive SATA directly. This is supported in the Litex infrastructure using the LiteSATA IP. Extra hardware is just a handful of capacitors and the SATA connectors. You still need a board like the TE0712 (the Ztex I use doesn't expose GTP pins). That was my intention for SBusFPGA under NetBSD, but is still theoretical (the TE0712 has been out of stock for months if not a couple of years, and isn't cheap). Probably less than 450€ for the assembled board, again the TE0712 is the expensive bit.
However in either case, one would need to write the appropriate drivers(s). Pure SATA might be easier, but is still a fairly complex piece of software that may or may not fit in the SCSI Manager framework. SATA on PCIe is even worse, as in addition to SATA one would need configuration and enumeration of the PCIe domain.
The hardware is quite feasible, the software would require a LOT of efforts.