DWDS/Dialing Concordance Code

+ CQueryWith*Node optimizations (actually previous check-in)
+ next: begin cleanup of debug code

+ CQuerySequenceNode and coefficient-based optimization: lots of debug code still in place, but looks solid

+ CQuerySequenceNode::EvaluateWithoutHits() optimization -- looking better

* binary search for CQuerySequenceNode -- not much joy (max ca. 3% faster)

* added macro DDC_GETBREAK_LB_MIN to QueryNode.cpp
- threshold for use of std::lower_bound() in break-identification step of CQueryNode::ConvertOccurrencesToHits()
- default value = 16k -> ca. 10% faster queries

v2.1.19-pre1 * added macro DDC_HIT_INTERSECTION_BINSEARCH to QueryNode.cpp
- use std::lower_bound for evaluating hit-list intersection and difference
- may be 3-5% faster for large hit-lists (function words)
* attempted profiling of CQueryNode::ConvertOccurrencesToHits() [perf, callgrind]: no insights

v2.1.18 release

+ force LC_NUMERIC=C in ConcordOptions.cpp (otherwise JSON responses may not be parseable)

+ v2.1.18-rc2: force LC_NUMERIC=C (in progress)

* fixed ddc_split in presence of legacy anonymized tokens (those beginning with -#)
- workaround (hack): all types with 0 occurrences in source corpus are assumed to by anonymized,
and are added to the output vocabulary of *every* partition -- this is a bit bloated but should work
* added ddc_index option --anonymize (-a): enable legacy anonymized tokens (disabled by default)
- fixes mantis #31586