[tmva][sofie] fix double-counted dilation in Conv im2col and adding regression test

[harz05]

Changes or fixes:

Conv generated invalid code and crashed (segfault) for dilation > 1. The operator builds a dilated _f weight buffer and expands fAttrKernelShape to the effective kernel size, but the im2col call was still passed the real dilation, so dilation was applied twice (expanded kernel + dilation). This produced a negative output dimension and an out-of-bounds write.

Fix: after the dilated _f is built, the dense im2col must use dilation 1, since the dilation is already folded into the expanded kernel shape and the _f layout. This is a no-op for dilation 1, so existing Conv tests are unaffected.

Also adds a ConvWithDilation test (3x3 kernel, dilation 2) with reference output. The suite had no dilation > 1 case before, which is why this was never caught. The full SOFIE test suite passes locally on master.

Checklist:

• tested changes locally
• updated the docs (if necessary)

This PR fixes #22473

[harz05]

kindly review @guitargeek ,thanks

[github-actions]

Test Results

    22 files      22 suites   3d 10h 2m 9s ⏱️
 3 853 tests  3 853 ✅ 0 💤 0 ❌
76 127 runs  76 127 ✅ 0 💤 0 ❌

Results for commit a6c3952.

♻️ This comment has been updated with latest results.

[harz05]

can we try a clean build once? @guitargeek

[harz05]

tests are passing, only clang formatting is failing