[ETHOSN] Improve inferring new shape of the Reshape operator

[NicolaLancellotti]

Fixes the case when reshape is > 4 dims. While this cannot be offloaded to the NPU, the check was previously producing an error preventing further compilation. The correct behavior is to ensure the check returns False and not offload the reshape.

cc @lhutton1

[NicolaLancellotti]

@lhutton1 @leandron

[lhutton1]

Thanks for the fix @NicolaLancellotti. Presumably due to this change: https://github.com/apache/tvm/pull/12594/files#diff-0cc1c5de01e0cebda9c447e34492ede3ae3121a9b6e11296218ecae47b299c68L329, the CI is failing on a network test. I also left a small comment below that you might as well add in at the same time :) Otherwise LGTM!

[lhutton1]

After reading apache/tvm-rfcs#88 I think we should quickly update the PR body to be more specific about the fix, WDYT? (No need to push a new commit, I can change the commit message so it matches the PR)

[NicolaLancellotti]

After reading apache/tvm-rfcs#88 I think we should quickly update the PR body to be more specific about the fix, WDYT? (No need to push a new commit, I can change the commit message so it matches the PR)

It is good to me to update the PR body.

[lhutton1]

LGTM!

[lhutton1]

Thanks @NicolaLancellotti!