Enabled multivariate blocks

[grishabh147]

I have made the following changes/additions in the code:

• Enabled multivariate blocks
• Enable multiple constraints in a block
• Used hypergraph abstraction to store the BlockNLPModel structure
• Added a test problem based on MPC for a chemical reactor
• Added docstrings

I am having some troubles with IPOPT on my computer and haven't been able to run full tests on this code. However, I have tested individual functions and they seem to be generating correct output. I will post an update whenever I am able to resolve the IPOPT issue.

My plan now is to also implement ADMM before we merge this to the main branch.

[grishabh147]

The latest commit brings the following updates:

• Added tests that are exclusive for BlockNLPModels.jl
• Made the changes requested by @frapac
• Removed dual_decomposition.jl and admm.jl from this repo. I will create a new repo for BlockNLPAlgorithms.jl and put these files there.

As you will see in simple_block_model.jl, even MadNLP.jl is not able to solve the full-space model created by FullSpaceModel(). I think there is some bug in my code that I am not able to figure out. Looking at the stacktrace makes me think that the issue is related to the size of sparse Hessian matrix that my code is generating. Can someone please help me debug this issue?