XML Pull Parser optimizations

This patch includes three improvements to the pull parser.

1. readValue now only uses the stringPool for all-whitespace strings.
   Values are generally not repeated, and this function can
   thrash the string pool in some cases

2. Added a fast path for the normal, correct, end-tag case.
   This avoids a hash and lookup in the stringPool

3. While parsing an element (not attribute) name, the namespace and
   post-namespace strings are generated using the stringPool.
   This saves the cost of calling String.substring in adjustNsp.

CRs-Fixed: 2128859

Change-Id: I69482a49bf52cb72d48bcbecd30fb56effe6a4ef
1 file changed