allow users to load custom hmm model

[name1e5s]

close #48

[codecov]

Codecov Report

Merging #92 (64158fc) into main (38fe280) will increase coverage by 1.17%.
The diff coverage is 52.63%.

@@            Coverage Diff             @@
##             main      #92      +/-   ##
==========================================
+ Coverage   86.41%   87.59%   +1.17%     
==========================================
  Files           8        7       -1     
  Lines        1067      693     -374     
==========================================
- Hits          922      607     -315     
+ Misses        145       86      -59     

Impacted Files	Coverage Δ
src/lib.rs	92.56% <ø> (+6.74%)	⬆️
src/hmm.rs	81.95% <52.63%> (-12.79%)	⬇️
src/keywords/tfidf.rs	83.82% <0.00%> (-3.82%)	⬇️
src/keywords/textrank.rs	84.09% <0.00%> (-2.83%)	⬇️
src/sparse_dag.rs	100.00% <0.00%> (ø)
tests/test_wasm.rs

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[messense]

I don't believe this is the best way to do it. How about making viterbi a method of HmmModel and let Jieba own a HmmModel?

struct Jieba {
    hmm_model: Option<HmmModel>,
    ...
}

impl Jieba {
    pub fn set_hmm_model(&mut self, hmm_model: HmmModel) -> &mut Self {
        ...
    }
}

struct HmmModel {
    ...
}

impl HmmModel {
    pub fn new(..) -> Self {
        todo!()
    }
}

#[cfg(feature = "default-hmm-model")]
impl Default for HmmModel {
    fn default() -> HmmModel {
        // load default hmm model
    }
}

[messense]

There is no point in use both lazy_static and once_cell, please also replace lazy_static with once_cell

[messense]

This would need a benchmark result to justify it.

[awong-dev]

I'm wondering if instead HMMModel should be a seprate struct and that Jieba's interface should be expanded to be

cut()
hmm_cut()

Will discussion back in #48