cpu: make use of cortexm.ld

[benpicco]

Contribution description

Some MCU families carry their own linker script when they can just as well make use of the generic one.

Testing procedure

Binaries should not change.

Issues/PRs references

split off #18608

[maribu]

looks good to me. I can test for nrf5x and sam, but not for other MCUs. I have the feeling that linker script changes are better tested before merging

[benpicco]

I can confirm that cpu/lm4f120 (ek-lm4f120xl) still works - but really it should suffice to compare the binaries for this, there should be no change.

[maribu]

Indeed, if binaries don't change there is no sense in testing.

For testing I ran

git checkout <PR>
for board in ek-lm4f120xl mbed_lpc1768 nrf51dk nrf52840dk nrf9160dk arduino-due; do make RIOT_CI_BUILD=1 BOARD=$board -C examples/default -j; done
mv examples/default/bin ~/bin-pr
git checkout master
for board in ek-lm4f120xl mbed_lpc1768 nrf51dk nrf52840dk nrf9160dk arduino-due; do make RIOT_CI_BUILD=1 BOARD=$board -C examples/default -j; done
mv examples/default/bin ~/
for board in ek-lm4f120xl mbed_lpc1768 nrf51dk nrf52840dk nrf9160dk arduino-due; do elf_diff --bin_prefix arm-none-eabi- --html_file ~/$board.html ~/bin-master/$board/default.elf ~/bin-pr/$board/default.elf; done
for board in ek-lm4f120xl mbed_lpc1768 nrf51dk nrf52840dk nrf9160dk arduino-due; do xdg-open ~/$board.html; done

The resulted report shows that in any case the exact same set of symbols was linked in and that the size of the symbols didn't change. I cannot rule out that due to different memory layout still issues arise, but I don't think the chance for such issues is high enough to justify the effort of testing them all.