The best place to gain an understanding of clone libraries (and therefore terms like library, library insert size, paired-end read and so on) is an introductory molecular biology textbook.

The library name will be an identifier string for the clone library that was sequenced. Insert size is the length in bases of a cloned insert in the library. What this really means depends on the context; it could be the precise size of a single insert, or it could be the desired size of insert aimed for during preparation in the laboratory. In reality, there will be a range of insert sizes in the library population. This brings us to insert size standard deviation, which is where the insert sizes are taken to follow a normal (bell curve) distribution - this is the standard deviation of that distribution.

You don't say what software you are using to convert SFF to Fasta, so I can't say how these values relate to your conversion task.

As to your specific question about method of SFF to Fasta conversion, the SFF specification describes the structure of the binary file. SFF data contains a sequence of estimates of homopolymer run lengths, plus basecalled nucleotide sequences and base quality values. Converting to Fasta means discarding all data except read name and basecalled nucleotide sequence.

Here is a post from the Biohaskell blog illustrating this nicely.