Reduce calls to std::random_device:

Existing per-thread PRNGs are individually initialized using calls
to std::random_device.

If merged, this commit will use a single PRNG, initialized from
std::random_device on startup, to seed the thread-specific PRNGs.

Acknowledgements:
    Thomas Snider, who reported this issue to Ripple on April 8, 2020.

src/ripple/basics/random.h

@@ -21,13 +21,13 @@
 #define RIPPLE_BASICS_RANDOM_H_INCLUDED
 
 #include <ripple/beast/xor_shift_engine.h>
-#include <boost/thread/tss.hpp>
 #include <cassert>
 #include <cstddef>
 #include <cstdint>
 #include <cstring>
 #include <random>
 #include <limits>
+#include <mutex>
 #include <ripple/beast/cxx17/type_traits.h> // <type_traits>
 
 namespace ripple {
@@ -67,27 +67,30 @@ inline
 beast::xor_shift_engine&
 default_prng ()
 {
-    static
-    boost::thread_specific_ptr<beast::xor_shift_engine> engine;
-
-    if (!engine.get())
+    // This is used to seed the thread-specific PRNGs on demand
+    static beast::xor_shift_engine seeder = []
     {
         std::random_device rng;
+        std::uniform_int_distribution<std::uint64_t> distribution{1};
+        return beast::xor_shift_engine(distribution(rng));
+    }();
 
-        std::uint64_t seed = rng();
+    // This protects the seeder
+    static std::mutex m;
 
-        for (int i = 0; i < 6; ++i)
+    // The thread-specific PRNGs:
+    thread_local beast::xor_shift_engine engine = []
+    {
+        std::uint64_t seed;
         {
-            if (seed == 0)
-                seed = rng();
-
-            seed ^= (seed << (7 - i)) * rng();
+            std::lock_guard lk(m);
+            std::uniform_int_distribution<std::uint64_t> distribution{1};
+            seed = distribution(seeder);
         }
+        return beast::xor_shift_engine{seed};
+    }();
 
-        engine.reset (new beast::xor_shift_engine (seed));
-    }
-
-    return *engine;
+    return engine;
 }
 
 /** Return a uniformly distributed random integer.
@@ -100,8 +103,8 @@ default_prng ()
 
     The randomness is generated by the specified engine (or
     the default engine if one is not specified). The result
-    is only cryptographicallys secure if the PRNG engine is
-    cryptographically secure.
+    is cryptographically secure only when the engine passed
+    into the function is cryptographically secure.
 
     @note The range is always a closed interval, so calling
           rand_int(-5, 15) can return any integer in the