CMake: preserve absolute pkg-config install directories

[fhgffy]

Problem

When CMAKE_INSTALL_LIBDIR or CMAKE_INSTALL_INCLUDEDIR is absolute, CMake installs to that directory but wavpack.pc prepends ${prefix}/. For example, prefix /tmp/wv-prefix and libdir /tmp/wv-lib incorrectly produce /tmp/wv-prefix//tmp/wv-lib, so downstream pkg-config consumers cannot find the installed files.

Change

Keep absolute values when generating the .pc file and retain the existing ${prefix} expressions for relative values. The final diff contains only six conditional lines in CMakeLists.txt, following the review request to omit the configuration test and its CI/distribution wiring.

Validation

After the scope reduction, five local CMake configuration cases pass: default, custom relative, absolute library only, absolute include only, and both absolute directories. All ten generated lib/include fields match the actual configured paths. git diff --check passes, and the aggregate diff against b6485a3 contains only CMakeLists.txt (+6/-0).

Earlier validation of this unchanged CMake production change included installed static consumers, relative prefix overrides, the pkg-config-module-off option, and the enabled program/CTest builds. After squashing, a fresh Linux Debug build with programs and BUILD_TESTING enabled also passes, and CTest passes wvtest (1/1). The full source tree is identical to the previous three-commit head. The regression is retained privately for verification and is no longer part of this PR. Windows and macOS were not run locally; current remote CI is separate.

[fhgffy]

Fixed the regression in 14a9aed: its default build directory now sits outside the source tree, and CMake cache entries are read as UTF-8. The original root-directory command, cmake -P cmake/tests/pkgconfig-paths.cmake, passes all five configuration cases with CMake 3.28.3 on Linux, including the UTF-8 paths in this workspace.

[dbry]

Thanks for the contribution!

Unfortunately I'm not that familiar with the cmake stuff (other than using it and basic edits) so I'd like to loop in the authors @sezero and @madebr . If they have no comments I'm happy to pull this in.

[sezero]

The change to CMakeLists.txt looks fair to me.

@madebr should look at the whole thing if he's able.

[madebr]

This step runs CMake configuration again.

I think this does not need a test and only the CMakeLists.text change should be kept.

[fhgffy]

Removed the configuration test and its CI/dist wiring in f463138. The PR now changes only the six lines in CMakeLists.txt. I kept the regression locally and reran the five directory configurations; all ten generated path fields pass.

[sezero]

Please squash your commits into a single one.

[fhgffy]

Would a squash merge on your side work for keeping this as one commit? The final diff is now just the six-line CMakeLists.txt change.

[fhgffy]

Squashed the three commits into f1690e0. The diff is unchanged: only six added lines in CMakeLists.txt. I reran all five pkg-config directory configurations, the Debug build, and CTest (wvtest); all passed.