Upgrade OpenFST to the latest C++14 version

I'm trying to run through the mini_librispeech/s5 example to get a better understanding on how to build models and have encountered an error. I built Kaldi using the MKL math library, and I'm seeing this issue in the relevant log:

```
# lattice-lmrescore --lm-scale=-1.0 "ark:gunzip -c exp/tri3b/decode_tgsmall_dev_clean_2/lat.1.gz|" "fstproject --project_output=true data/lang_test_tgsmall/G.fst |" ark:- | lattice-lmrescore-const-arpa --lm-scale=1.0 ark:- data/lang_test_tglarge/G.carpa "ark,t:|gzip -c>exp/tri3b/decode_tglarge_dev_clean_2/lat.1.gz"
# Started at Tue Jun 15 00:04:31 UTC 2021
#
lattice-lmrescore --lm-scale=-1.0 'ark:gunzip -c exp/tri3b/decode_tgsmall_dev_clean_2/lat.1.gz|' 'fstproject --project_output=true data/lang_test_tgsmall/G.fst |' ark:-
lattice-lmrescore-const-arpa --lm-scale=1.0 ark:- data/lang_test_tglarge/G.carpa 'ark,t:|gzip -c>exp/tri3b/decode_tglarge_dev_clean_2/lat.1.gz'
FATAL: SetFlags: Bad option: --project_output=true
ERROR: FstHeader::Read: Bad FST header: fstproject --project_output=true data/lang_test_tgsmall/G.fst |. Magic number not matched. Got: 0
ERROR (lattice-lmrescore[5.5.873~1-75ec]:ReadFstKaldi():kaldi-fst-io.cc:35) Reading FST: error reading FST header from 'fstproject --project_output=true data/lang_test_tgsmall/G.fst |'

[ Stack-Trace: ]
lattice-lmrescore(kaldi::MessageLogger::LogMessage() const+0xa18) [0x5463b2]
lattice-lmrescore(kaldi::MessageLogger::LogAndThrow::operator=(kaldi::MessageLogger const&)+0x11) [0x47da63]
lattice-lmrescore(fst::ReadFstKaldi(std::__cxx11::basic_string, std::allocator >)+0x1af) [0x4d012f]
lattice-lmrescore(main+0x20c) [0x478cf3]
/lib64/libc.so.6(__libc_start_main+0xea) [0x7fe66a53b0ba]
lattice-lmrescore(_start+0x2a) [0x478a4a]

WARNING (lattice-lmrescore[5.5.873~1-75ec]:Close():kaldi-io.cc:515) Pipe fstproject --project_output=true data/lang_test_tgsmall/G.fst | had nonzero return status 256
kaldi::KaldiFatalErrorLOG (lattice-lmrescore-const-arpa[5.5.873~1-75ec]:main():lattice-lmrescore-const-arpa.cc:117) Done 0 lattices, failed for 0
# Accounting: time=0 threads=1
# Ended (code 1) at Tue Jun 15 00:04:31 UTC 2021, elapsed time 0 seconds
exp/tri3b/decode_tglarge_dev_clean_2/log/rescorelm.1.log (END)
```

All of the log files of the form exp/tri3b/decode_tglarge_dev_clean_2/log/rescorelm.*.log seem to have the same error.

It seems to be a bad command,

`lattice-lmrescore --lm-scale=-1.0 "ark:gunzip -c exp/tri3b/decode_tgsmall_dev_clean_2/lat.10.gz|" "fstproject --project_output=true data/lang_test_tgsmall/G.fst |" ark:- | lattice-lmrescore-const-arpa --lm-scale=1.0 ark:- data/lang_test_tglarge/G.carpa "ark,t:|gzip -c>exp/tri3b/decode_tglarge_dev_clean_2/lat.10.gz"`

when I run the command by itself (after running `source path.sh`) outside of run.sh, it throws the same error.

I don't understand the format of the command enough to correct it myself, nor how to fix it in the script.