mpl2: post-annealing step to push macros to boundaries

[AcKoucher]

#1930 showed that we actually don't allow different x & y coordinates for macro placement halos in mpl2, so #1933 was opened.

When I was testing the distinct x & y halos for mock-array's test case used in #4085, two problems came up:

• The macros' locations inside the array were wrong (should be fixed by #4976)
• The array was getting split in two as mpl2 was trying to push the macros to the boundaries by a post-annealing function.

This PR is to change the alignHardMacroGlobal function by a new step in mpl2 that happens before we improve orientation. Basically, we push macros that are close enough to the boundaries if:

1. They don't overlap with blockages generated for Pin Access
2. They don't overlap with other macros
3. Pushing won't destroy the shape of an array of macros.

[AcKoucher]

It looks like there's something going on with clang-tidy.

[maliberty]

secure ci?

[AcKoucher]

I started a secure-ci

[maliberty]

@osamahammad21 this comes with one new drc in ca53 but it seems to be just a case of perturbing things happens to trigger a routing issue. I don't see a problem in the macro placement so I'll ask you to look at it but don't want to hold this PR for it.