Workaround for "PHC time jumps ~4 seconds there and back on AQC107"

[peci1]

Workarouds issue https://www.spinics.net/lists/netdev/msg1013570.html .

Summary: PHC time jumps ~4 seconds there and back on AQC107

I'm testing on multiple copper-PHY NICs with AQC107 and the problem is always the same:

• QNAP QXG-10G2TB, PCIe, FW 3.1.121
• Sanlink3 N1, Thunderbolt, FW 3.1.121
• OWC OWCTB3ADP10GBE, Thunderbolt, FW 3.1.106 (it cannot run 3.1.121)
• Sonnettech Solo10G 10GBASE-T, Thunderbolt, FW 3.1.121

I run the tests on multiple PCs with different CPU archs and the problem is always the same:

• Intel NUC (Thunderbolt, x86-64, kernel 6.8, upstream atlantic driver)
• NVidia Jetson (PCIe, aarch64, kernel 4.9.337 + out-of-tree driver https://github.com/Aquantia/AQtion ).

When I enable PTP on the card (e.g. hwstamp_ctl -i eth2 -t 1) and run phc2sys from linuxptp
to sync the card's PHC to another clock (AQ107 is time sink), I get time jumps of the PHC clock
of ~4 secs in the future and then back (after ~5 secs). These jumps happen in random intervals
and I haven't found a pattern. Usually, it happens every 20-100 secs, but there have been even
much longer intervals. These time jumps are corrected by phc2sys quite quickly (within ~10 seconds),
but during this interval, the time is wrong.

As concluded in the discussion on netdev, this PR brings in a workaround that seems to work and seems to have no side-effects. But it doesn't solve the root cause.