Write some benchmarks for the different sub-crates

[angelip2303]

Alongside with the lost on performance, I think that it would be fantastic to have some sort of benchmarking bundled withing the CI pipeline. Hence, we could see if there is any performance lost from release to release. The current benchmarking is performed manually, which is not handy to use :/.

We should focus on macro-benchmarking, as I believe what we want to measure is the time it takes to perform actual operations, as end-users would do.

• Profiling
• Micro-benchmarking of rudof_lib
• Flamegraph
• CI

[labra]

One possible source of this regression in performance is the use of the RdfData struct which contains a copy of the in-memory [SRDFGraph] and a vector of endpoints.

I wonder if we could change that code to contain only a reference to the SRDFGraph changing the code to somathing like:

pub struct RdfData<'a> {
    /// Current focus node used when parsing
    focus: Option<OxTerm>,

    /// List of SPARQL endpoints
    endpoints: Vec<SRDFSparql>,

    /// In-memory graph
    graph: Option<&'a SRDFGraph>,

    /// In-memory Store used to access the graph using SPARQL queries
    store: Option<&'a Store>,
}

One problem is that the struct would require to carry the lifetime parameter, which makes it a bit less ergonomic to use in the python bindings...but I think that struct is not exported in the python bindings,. It would be interesting to try that change and see how it affects the performance because I think it would require less clones...