Don't synthesize 0 of the ArraySubscriptExpr type in forward mode

[PetroZarytskyi]

Currently, Clad can't differentiate ArraySubscriptExpr of non-numerical types. This happens because it attempts to synthesize a 0 of the same type, which is not always numerical. To avoid this error, we can use getZeroInit instead.
Here's a reproducer (it's added to the tests in this PR):

struct A {
  double mem;
  A(double p_mem = 0) : mem(p_mem) {}
};

double fn18(double i, double j) {
  A v[2] = {2, 3};
  v[0] = 9 * i;
  return v[0].mem;
}

Differentiating this in forward mode produces an error.

This bug arises when differentiating HepEmShow.

[github-actions]

clang-tidy made some suggestions

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 93.86%. Comparing base (216201a) to head (2b02d7c).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #960      +/-   ##
==========================================
+ Coverage   93.85%   93.86%   +0.01%     
==========================================
  Files          55       55              
  Lines        7892     7892              
==========================================
+ Hits         7407     7408       +1     
+ Misses        485      484       -1     

Files	Coverage Δ
lib/Differentiator/BaseForwardModeVisitor.cpp	98.95% <100.00%> (ø)

... and 1 file with indirect coverage changes

Files	Coverage Δ
lib/Differentiator/BaseForwardModeVisitor.cpp	98.95% <100.00%> (ø)

... and 1 file with indirect coverage changes

[github-actions]

clang-tidy review says "All clean, LGTM! 👍"

[vgvassilev]

LGTM!