Add @SQ-AN alternative reference sequence names

[jmarshall]

Proposed text to implement #100, "suggestion: aliases in @SQ section".

Also take the opportunity to pre-emptively restrict the characters allowed in these alternative names.

As noted in last September's meeting, restricting the characters allowed in AN alternative names could be a first step towards applying the same restriction in SN canonical sequence names and finally being able to parse the colon in chr:beg-end unambiguously.

There is debate to be had about which punctuation characters to forbid; there are existing points for consideration in #100, #124, and #167.

The primary motivation here is to disallow : so chr:beg-end region syntax is unambiguous, so we have taken advantage of colon being available as the delimiter, inspired by TAG:TYPE:VALUE delimiters.

Other motivation might be to disallow , so chr1,chr2,chr3 list syntax is unambiguous. If disallowing comma has particular support, perhaps using comma as the AN delimiter would be preferable.

[daviesrob]

For information, I did a count of the number of punctuation characters in the entry names of our set of 292 reference fasta files. I got:

 # 203
 % 203
 * 525
 + 1
 , 496
 - 154226
 . 1826561
 : 1577
 = 26
 _ 4961932
 | 1098333

The 496 commas all come from GRCh38.rRNA.fa which contains entries like this:

>ENSG00000276700|RNA, 5.8S ribosomal 5 [Source:HGNC Symbol;Acc:HGNC:37660]|RNA5-8S5|rRNA
>gi|189571632|ref|NR_023379.1| Homo sapiens RNA, 5S ribosomal 17 (RNA5S17), ribosomal RNA
>ENSG00000199334|RNA, 5S ribosomal 11 [Source:HGNC Symbol;Acc:HGNC:34372]|RNA5S11|rRNA
>ENSG00000201966|RNA, 5.8S ribosomal pseudogene 4 [Source:HGNC Symbol;Acc:HGNC:41958]|RNA5-8SP4|rRNA

Colons are in Homo_sapiens.GRCh38_full_analysis_set_plus_decoy_hla.fa (HLA regions) ...:

>HLA-A*01:01:01:01	HLA00001 3503 bp
>HLA-A*01:01:01:02N	HLA02169 3291 bp
>HLA-A*01:01:38L	HLA03587 3374 bp
>HLA-A*01:02	HLA00002 3374 bp

... and a few other files:

Plasmodium_berghei.rRNA.fa:
>berg10:rRNA:rfamscan:938985-939103

Plasmodium_falciparum.rRNA.fa:
>MAL14_5S_2|MAL14_5S_2:rRNA

Trichuris_muris_v0_1_281111.fa:
>NODE_48597_length_3376_cov_5.922986.1.1633:00009.7180000457210-00010.7180000828236

Trichuris_muris_11_03_2015.fa:
>U_TMUE_mito:1

haemonchus_V1.fa:
>scaffold2.1_size831208:1-30808

I guess GRCh38 is the biggest problem here.

[jkbonfield]

Inspired by TAG:TYPE:VALUE is OK, but note we already have TAG:TYPE:VALUE,VALUE,VALUE for some tags and TAG:TYPE:VALUE;VALUE;VALUE for others when we're dealing with lists of values. Frankly anything you do here will fit in given lack of consistency elsewhere!

Disallowing colon seems very logical and solves a lot of problems we've had, but I'd be inclined to still use comma as the separator. I don't have any strong feelings though so would accept it as it is or with comma instead.

[jmarshall]

Recreated using commas as a more natural list delimiter. I hear there is either VCF or at least bcftools option notation using commas as refseq name delimiters, and, closer to home, SAM's SA tag already uses a comma to terminate a reference sequence name.

[jmarshall]

More previous discussion of forbidding characters in #124 and samtools/samtools#149.