Gensim

* added `corpora.IndexedCorpus`, a base class for corpus serializers (thx to Dieter Plaetinck). This allows corpus formats that inherit from it (MmCorpus, SvmLightCorpus, BleiCorpus etc.) to retrieve individual documents by their id in O(1), e.g. `corpus[14]` returns document 14.
* merged new code from the LarKC.eu team (`corpora.textcorpus`, `models.logentropy_model`, lots of unit tests etc.)
* fixed a bug in `lda[bow]` transformation (was returning gamma distribution instead of theta). LDA model generation was not affected, only transforming new vectors.
* several small fixes and documentation updates

* new LDA implementation after Hoffman et al.: Online Learning for Latent Dirichlet Allocation
* distributed LDA
* updated LDA docs (wiki experiments, distributed tutorial)
* matrixmarket header now uses capital 'M's: MatrixMarket. (André Lynum reported than Matlab has trouble processing the lowercase version)
* moved code to github
* started gensim Google group

* added workaround for a bug in numpy: pickling a fortran-order array (e.g. LSA model) and then loading it back and using it results in segfault (thx to Brian Merrel)
* bundled a new version of ez_setup.py: old failed with Python2.6 when setuptools were missing (thx to Alan Salmoni).

* further optimization to LSA; this is the version used in my NIPS workshop paper
* got rid of SVDLIBC dependency (one-pass LSA now uses stochastic algo for base-base decompositions)

* sped up Latent Dirichlet ~10x (through scipy.weave, optional)
* finally, distributed LDA! scales almost linearly, but no tutorial yet. see the tutorial on distributed LSI, everything's completely analogous.
* several minor fixes and improvements; one nasty bug fixed (lsi[corpus] didn't work; thx to Danilo Spinelli)

* added stochastic SVD decomposition (faster than the current one-pass LSI algo, but needs two passes over the input corpus)
* published gensim on mloss.org