Refactor pgp_encrypted_material_t to C++ classes.

[ni4]

Just few more types left to avoid unions and move pgp::mpi to std::vector.

[codecov]

Codecov Report

Attention: Patch coverage is 86.46154% with 44 lines in your changes missing coverage. Please review.

Project coverage is 85.14%. Comparing base (884b52b) to head (39e3e5e).

Files with missing lines	Patch %	Lines
src/lib/key_material.cpp	88.60%	9 Missing ⚠️
src/librepgp/stream-dump.cpp	87.32%	9 Missing ⚠️
src/lib/rnp.cpp	27.27%	8 Missing ⚠️
src/lib/sig_material.cpp	77.41%	7 Missing ⚠️
src/lib/enc_material.cpp	84.21%	6 Missing ⚠️
src/librepgp/stream-packet.cpp	85.71%	2 Missing ⚠️
src/lib/crypto/signatures.cpp	85.71%	1 Missing ⚠️
src/lib/sig_material.hpp	83.33%	1 Missing ⚠️
src/librepgp/stream-parse.cpp	85.71%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2304      +/-   ##
==========================================
+ Coverage   84.90%   85.14%   +0.24%     
==========================================
  Files         114      118       +4     
  Lines       22788    22780       -8     
==========================================
+ Hits        19349    19397      +48     
+ Misses       3439     3383      -56     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[maxirmx]

There are multiple constructs where a set of dynamic_cast's in placed inside case statement. I marked it in rnp.cpp but I saw it in stream-dump.cpp and other files

While not an error, such pattern (dynamic_cast's inside case) clearly indicate an option to create a virtual function.

May be you can consider it fot future refactoring

[maxirmx]

IMHO this piece of code call for virtual function

something like

virtual mpi* SigMaterial::get_mpi() = 0;

and appropriate redefinitions for EgSigMaterial et al

[ni4]

I don't like the way it is currently done as well. The problem here is that different signature types may have different MPIs. In this case it is simple, one or two, but for EncMaterial and KeyMaterial it is more complicated, however I think we should use the same approach in both cases. Don't know a good solution yet.