massively sped up combineTCR with Rcpp

[Qile0317]

combineTCR is one of the most important functions but it's kinda slow, running originally in ~ 24 seconds for contig_list on my device. I used the profvis profiling library to see the speed bottlenecks, and see that .parseTCR is the main time consumer. Looking deeper into the function, this PR solves one of them - finding the indicies of barcodes of Con.df$barcode of length N that are in data2$barcode of length M, by constructing a barcode index in O(N+M) time in Rcpp. This has cut the runtime down to ~17 seconds. I'm working on fully re-implementing .parseTCR in Rcpp atm to see if theres another time increase.

[Qile0317]

a new rcppConstructBarcodeIndex function was the main addition, which def should be able to be used by combineBCR as well but that needs some testing first before the transition to ensure nothing breaks.

[Qile0317]

Just implemented the parseTCR step fully in Rcpp but there are some critical errors I'm trying to fix. Although running the combineContigs test file seems faster now so thats a good sign

[Qile0317]

Ok, with the latest commit the segmentation fault cause has been figured out. I've added one small testcase but for the most part I believe the implementation is completely correct as past testcases on combineTCR are still passing. And the runtime for combineTCR has been massively reduced from 17 seconds to about 0.5 seconds! This can also be pretty easily implemented to combineBCR but testdata is needed.

[Qile0317]

With the latest commit 50ac737 I just re-implemented the barcode indexing for parseBCR temporarily, but testing is definetely needed, and the testdata should be generated with the old version. If you think this is unsound feel free to undo all the changes related to .parseBCR but otherwise I still advocate for the combineTCR changes