Merge bitcoin/bitcoin#22923: test: Switch multiprocess to clang i686 build

fa309ee61c bench: Fix 32-bit compilation failure in addrman bench (MarcoFalke)
fae0295a79 ci: Switch multiprocess to i686 build (MarcoFalke)

Pull request description:

  Building for i686 with clang helps to catch bugs early for:
  * The OSS-Fuzz i686 clang libFuzzer build
  * The arm 32-bit native clang build

  Fixes  #22889

ACKs for top commit:
  hebasto:
    ACK fa309ee61c

Tree-SHA512: 581820d319aae2fcd4dd44979ee3d4164a575f0438476890aa2a7447f1392a5da26766cd6ab954530499b54f66eec2417bdeefdd7efb19bc27dd679cd2b9d0ce

src/bench/addrman.cpp

@@ -5,6 +5,7 @@
 #include <addrman.h>
 #include <bench/bench.h>
 #include <random.h>
+#include <util/check.h>
 #include <util/time.h>
 
 #include <optional>
@@ -110,7 +111,8 @@ static void AddrManGood(benchmark::Bench& bench)
      * we want to do the same amount of work in every loop iteration. */
 
     bench.epochs(5).epochIterations(1);
-    const size_t addrman_count{bench.epochs() * bench.epochIterations()};
+    const uint64_t addrman_count{bench.epochs() * bench.epochIterations()};
+    Assert(addrman_count == 5U);
 
     std::vector<std::unique_ptr<CAddrMan>> addrmans(addrman_count);
     for (size_t i{0}; i < addrman_count; ++i) {