Rebrand hush3 to DragonX and share RandomX dataset across mining threads

Minimal rebrand (see compliant-rebrand branch for full rebrand):
- Rename binaries: hushd/hush-cli/hush-tx → dragonxd/dragonx-cli/dragonx-tx
- Default to DRAGONX chain params without -ac_* flags (randomx, blocktime=36, private=1)
- Update configure.ac: AC_INIT([DragonX],[1.0.0])
- Update client version string and user-agent to /DragonX:1.0.0/
- Add chainparams.cpp with DRAGONX network parameters
- Update build.sh, miner.cpp, pow.cpp for DragonX
- Add bootstrap-dragonx.sh utility script
- Update .gitignore for release directory

Share single RandomX dataset across all mining threads:
- Add RandomXDatasetManager with readers-writer lock, reducing RAM from
  ~2GB per thread to ~2GB total plus ~2MB per thread for the VM scratchpad
- Add LogProcessMemory() diagnostic helper for Linux and Windows

src/clientversion.cpp

@@ -32,7 +32,7 @@
  * for both bitcoind and bitcoin-core, to make it harder for attackers to
  * target servers or GUI users specifically.
  */
-const std::string CLIENT_NAME = GetArg("-clientname", "GoldenSandtrout");
+const std::string CLIENT_NAME = GetArg("-clientname", "DragonX");
 
 /**
  * Client version number