fix: enable use of point features

[m-jahn]

• currently point features such as externally annotated transcription start or stop sites can be supplied
• however these are never used by the program, see e.g. Point features seem not to be used when predicting transcripts with option --ptf #466
• it is also not clear how and when they are supposed to be used, see Point features seem not to be used when predicting transcripts with option --ptf #466 How to integrate point-feature option in the standard workflow ? #339 -ptf option - recommended input & implementation details? #330
• after some digging why point features can be supplied but are never used, I found a bug in the code snippet that adds point features to the current bundle
• now, point features are correctly used to refine transcript (not exon) boundaries, as long as they are within the predicted transcript boundary (!)

There are at least two open tasks that remain to make this feature fully functional @gpertea .
Can you please look into the following and maybe add a commit to this PR? Would be fantastic.

Open Task 1: there is currently no parameter to control the usage of point features, i.e. a tolerance in bp until a PTF is used or not

Open Task 2: I could absolutely not find out why point features that exceed the transcript boundary even by just 1 bp are ignored?! The answer must be hidden in the 20,000 lines rlink.cpp file. The sheer size and convolution of this code base is unfortunately too difficult to handle. I tried to comment out some of the filters but had no success.

The point features can be tested using the test data like this:

./stringtie -L -v --ptf tests/long_reads_ptf.tsv -o tests/long_reads.out.gtf tests/long_reads.bam

where long_reads_ptf.tsv is a TSV table like:

chr19	567222	-	TSS
chr19	571735	-	CPAS
chr19	571311	+	TSS
chr19	583492	+	CPAS