Airport/city codes should take the precedence over alternate names

[da115115]

When seaching for KHI, the city of Jakarta, Indonesia (ID), is returned, whereas KHI is the code of Karachi, Pakistan (PK).

The cause is that:

1. The Hakka Chinese translation of Jakarta is "Ngâ-kâ-tha̍t Sú-tû Thi̍t-khî", including the "khî" keyword, which is therefore part of the indexing keywords for Jakarta city.
2. The respective PageRank values of KHI and JKT/CGK are 8% and 36.5%. Hence, KHI will match almost exactly (99.99%) with both Karachi and Jakarta. With the PageRank values, Jakarta comes out with an overall matching weight of ~36% (compared to the ~8% of Karachi).

[da115115]

When 3-letter/4-letter codes are entered, and they those match with a IATA, ICAO or FAA code, the corresponding POR entries should be selected. One way to do that is to set the final matching weight (after application of PageRank) to 100%.

[da115115]

Though working quite nicely, there are still some imperfections, some of which may be simply fixed. More specifically, the http://search-travel.org/search/?q=sez+airport request yields "Mahé (SEZ) -- Vadsø Airport (VDS)", due to the fact that {'sez', 'airport'} better matches than {'sez airport'}, as 'sez' now matches with 100% and 'airport' matches with 2.45%. Several options in order to improve the algorithm:

• When 'airport' (or any other "back-listed" word such as 'city' or 'international') is the single word of the query, it should match at 0%.
• '<IATA/ICAO code> airport' or '<IATA/ICAO code> city' are also given a matching weight of 100%.

[da115115]

When indexing a term, the weight may be specified; see the documentation of Xapian::TermGenerator::index_text() function for more details. By default, a weight of 1 is applied. For IATA/ICAO/FAA codes, a weight of 2 may be applied. Note that, for now, the Place source code implements the indexation in a generic way through a generic list of terms stored in Place. We could therefore have a map of list of terms: the key would be the weight and the value would be the list of terms:

  typdef <const Weight_T, StringSet_T> TermSetMap_T;
  TermSetMap_T _termSetMap;

Then, in the Xapian indexing code, the terms would be indexed according to their weight.

[da115115]

Implemented as for the issue #5 (Indexing travel types with PageRank weights).