[SIT] Add post proc options for l2norm

[svigh]

Details:

• Add post proc options for l2norm

Tickets:

• EISW-215382

AI Assistance:

• AI assistance used: yes
• If yes, summarize how AI was used and what human validation was performed (build/tests/manual checks): implemented dequant logic, manually tested

[Copilot]

Pull request overview

This PR extends the Intel NPU single-image-test tool’s l2norm comparison mode with optional output dequantization (scale + zero-point) so quantized inference outputs can be converted before computing the L2 norm against reference tensors.

Changes:

• Added new CLI flags l2norm_dequant_scale and l2norm_dequant_zp (global or per-output) and surfaced them in tool help output.
• Updated l2norm computation path to optionally dequantize the inference output before comparison, leaving the reference tensor untouched.
• Added helper utilities to distinguish global vs explicit per-layer (per-output) values.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
src/plugins/intel_npu/tools/single-image-test/README.md	Documents new l2norm dequantization flags.
src/plugins/intel_npu/tools/single-image-test/main.cpp	Implements optional dequantization in l2norm metric evaluation.
src/plugins/intel_npu/tools/single-image-test/argument_parse_helpers.hpp	Declares new flags and adds helpers for global vs explicit per-layer values.
src/plugins/intel_npu/tools/single-image-test/argument_parse_helpers.cpp	Defines new flags, prints them in l2norm mode, and implements the new helper functions.