Enhance `provoc` with GLM and Flexible Data Handling

[danerkestey]

Refactored provoc to support formula-based inputs and user-defined lineages, improving data handling flexibility and alignment with binomial glm.

[danerkestey]

General:

• [x] Remove "Imports" Section:
• [x] Use provoc::: for any internal function calls within the source code.

Default Parameter:

• [x] Ensure the function can operate with by = NULL as the default parameter setting for better handling of multiple samples in a file

Function Optimization:

• [x] Replace glm() with provoc::provoc_optim() for fitting purposes. This change is essential to ensure estimates are positive, sum to less than 1 through constrained optimization, and use a linear link function instead of the logistic link function.

@examples Chunk:

• [x] Make sure the @examples section in docs can run in a clean environment, assuming the package is loaded. Include basic code applying this function to Baaijens as part of the examples.

Handling of mutation_defs:

• [x] Review the implementation of mutation_defs to ensure it matches mutation names in the data correctly. This involves checking the matrix dimensions are correctly indexed by mutation and variant names and utilizing provoc::fuse() or a better implementation for this purpose.
• [x] Implement a check to confirm that lineage names in mutation_defs are present to catch potential typos from the user.

Data and Lineage Names Handling:

• [x] Modify the code to not expect lineage names as columns in the input data unless mutation_defs is provided, considering most users won't have lineage-related columns in their original datasets -> met through fuse

Output Object Requirements:

• [x] Ensure the output object contains all necessary information for the predict.provoc(), to facilitate accurate predictions.

Parse_Mutations Improvements:

• [x] Initialize new_muts vector before loop
• [x] Turn new_muts into a dataframe