Ledger acquire fixes/cleanups/logging

* Inbound ledger and SHAMapAddNode cleanup
    * Log acquire stats
    * Fix progress logic
    * Remove ledgers we no longer need to acquire
    * Stash stale state data in our fetch pack, it can still be useful
    * Stash in fetch pack if acquire terminated while job was pending
    * Account for duplicate/invalid nodes in a few cases previously missed
    * Dispatch each InboundLedger once (not per data)
    * Trigger only the "best" peer
    * Don't call tryAdvance on failed acquires

src/ripple_app/misc/NetworkOPs.cpp

@@ -50,7 +50,7 @@ public:
         , mLastCloseConvergeTime (1000 * LEDGER_IDLE_INTERVAL)
         , mLastCloseTime (0)
         , mLastValidationTime (0)
-        , mFetchPack ("FetchPack", 2048, 20)
+        , mFetchPack ("FetchPack", 65536, 45)
         , mFetchSeq (0)
         , mLastLoadBase (256)
         , mLastLoadFactor (256)
@@ -3066,6 +3066,11 @@ int NetworkOPsImp::getFetchSize ()
 
 void NetworkOPsImp::gotFetchPack (bool progress, uint32 seq)
 {
+
+    // FIXME: Calling this function more than once will result in
+    // InboundLedgers::gotFetchPack being called more than once
+    // which is expensive. A flag should track whether we've already dispatched
+
     getApp().getJobQueue ().addJob (jtLEDGER_DATA, "gotFetchPack",
                                    BIND_TYPE (&InboundLedgers::gotFetchPack, &getApp().getInboundLedgers (), P_1));
 }