Here is the best documentation (that I know of) for TCGA SNP-array based CNV data. Regd. your two questions:
CBS segmentation algorithm identifies regions in the genome that, in spite of noise, probably have a uniform underlying copy number. The "segment mean" of each region is reported in the level 3 file, and can be used as the estimated CN-ratio for the segment.
"nocnv" just means that germline CN variations are removed. In TCGA, they ran SNP arrays on normal tissue too.