Limit default value of TEST_NP to at most 4 cores

When no default value is provided for TEST_NP, value `nproc / 2 + 1`
is used instead, which can be large on hyperthreaded CPUs. This leads
to issues in parallel tests that don't scale well with a large number
of cores (e.g. when there aren't enough particles) or with a large
odd number of cores (e.g. LB or P3M).

cmake/unit_test.cmake

@@ -2,6 +2,9 @@ if(NOT DEFINED TEST_NP)
   include(ProcessorCount)
   ProcessorCount(NP)
   math(EXPR TEST_NP "${NP}/2 + 1")
+  if(${TEST_NP} GREATER 4)
+    set(TEST_NP 4)
+  endif()
 endif()
 
 if(EXISTS ${MPIEXEC})