Fix performance regression when appending one vector to another

When appending a QVector to an existing vector the code would
unconditionally realloc the vector instead of first checking
whether we can do without. This gives a quadratic behaviour when
repeatedly appending a vector to another.

Change-Id: I2cd81253e6a8aec0bc4402e6fbda262435080966
Reviewed-by: Qt Doc Bot <qt_docbot@qt-project.org>
Reviewed-by: Simon Hausmann <simon.hausmann@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

src/corelib/tools/qvector.h

@@ -650,8 +650,12 @@ QVector<T> &QVector<T>::fill(const T &from, int asize)
 template <typename T>
 QVector<T> &QVector<T>::operator+=(const QVector &l)
 {
-    int newSize = d->size + l.d->size;
-    reallocData(d->size, newSize);
+    uint newSize = d->size + l.d->size;
+    const bool isTooSmall = newSize > d->alloc;
+    if (!isDetached() || isTooSmall) {
+        QArrayData::AllocationOptions opt(isTooSmall ? QArrayData::Grow : QArrayData::Default);
+        reallocData(d->size, isTooSmall ? newSize : d->alloc, opt);
+    }
 
     if (d->alloc) {
         T *w = d->begin() + newSize;