Revert #120 "Prevent variable nullification on later imports"

[m-dzianishchyts]

What does this PR do?

Reverts #120 and updates integration test cases reproducing the reported issues.

[andrus]

@m-dzianishchyts , @stariy95 : guys, I tried to understand how this PR is different from the original #120. So I manually reverted all #120 commits (already pushed to "main") and then did a merge --squash of this one on top. The only change that I see is introduction of KernelExecutionIT.java test. But the original problem is still there when I run it manually. It was not fixed.

[andrus]

Ok, I was able to rewrite KernelExecutionIT in a way that reproduces the issue. It is committed to main (see a @Disabled test). The reason the original test didn't fail was that each line was executed in a separate cell. Somehow combining things in cells in a certain way causes this problem. Here are Claude findings on which combos work and which don't:

  ┌──────────────────────────────────────────────────────────┬────────┐
  │                          cells                           │ result │
  ├──────────────────────────────────────────────────────────┼────────┤
  │ 5 separate cells (what the IT does)                      │ ✅ ok  │
  ├──────────────────────────────────────────────────────────┼────────┤
  │ all 5 in one cell                                        │ ❌ NPE │
  ├──────────────────────────────────────────────────────────┼────────┤
  │ %maven alone, then remaining 4 lines in one cell         │ ❌ NPE │
  ├──────────────────────────────────────────────────────────┼────────┤
  │ %maven / import+var om+import / om.getClass()            │ ❌ NPE │
  ├──────────────────────────────────────────────────────────┼────────┤
  │ %maven / import+var om / import+om.getClass()            │ ✅ ok  │
  └──────────────────────────────────────────────────────────┴────────┘

One little thing left - figure out how to fix it 🙂