Fix nav2_smac_planner robin_hood Windows build

[Tobias-Fischer]

This updates the kilted nav2_smac_planner Windows patch after the upstream Navigation2 review feedback on ros-navigation/navigation2#6276.

The first CI probe showed that removing RoboStack's std::unordered_map fallback exposed the actual Windows failure: MSVC could not parse the vendored robin_hood.h macro because a // NOLINT comment followed a macro-continuation backslash.

This PR now keeps robin_hood::unordered_node_map on Windows and applies the same macro-continuation fix proposed upstream, while keeping the nav2_smac_planner build-number bump so CI rebuilds the package.

Validation:

• Patch applies cleanly against the exact kilted release tag used by CI: release/kilted/nav2_smac_planner/1.4.2-1.
• No local build was run.

[Tobias-Fischer]

This probe did what we needed:

• Linux and macOS passed.
• Windows failed while compiling a_star.cpp.
• The first MSVC errors are in the vendored robin_hood.h macro body:

robin_hood.h(151,17): error C2059: syntax error: 'return'
robin_hood.h(152,67): error C2059: syntax error: 'constant'
robin_hood.h(153,13): error C2059: syntax error: '}'
robin_hood.h(153,13): error C2143: syntax error: missing ';' before '}'

The cause appears to be the // NOLINT comment after a macro-continuation backslash in ROBIN_HOOD_COUNT_TRAILING_ZEROES. I amended the upstream Navigation2 PR to fix that macro continuation instead of switching Windows to std::unordered_map: ros-navigation/navigation2#6276

[Tobias-Fischer]

Can we merge this @traversaro or could there be ABI issues?

[traversaro]

Can we merge this @traversaro or could there be ABI issues?

The header is installed, by the library is actually a plugin that is loaded at runtime, so I think it is actually safe to merge.