Add compatibility with arrow to deal with larger-than-memory data

[rafapereirabr]

Problem:
Oftentimes, travel cost matrices can get really big in large study areas. This creates a challenge of dealing with larger-than-memory data. The current version of {accessibility} cannot cope with this problem. Although the package is super efficient because it is built on top od {data.table}, the package still requires users to load the travel_matrix and land_use_data input to RAM memory, which can be a constraint for these larger-than-memory cases.

Proposed solution:
One solution to this would be to make the {accessibility} compatible with {arrow} (link). There are different ways to do this. The alternative I would suggest would use the following workfkow:

• If the user passes travel_matrix and land_use_data inputs as data.tables or data.frames, then the package follows the current natural flow.
• If the user passes a string path pointing to the location of .parquet or .csv files that contain the travel_matrix and land_use_data inputs, then the package follows a different flow to calculate the indicador using a {dplyr} + {arrow} syntax

Implications:

• For users, this only means more flexibility to work with larger-than-memory data and greater computation speed.
• For developers, this means we would need to make a few changes to check inputs, but most of the work would be translating the accessibility functions into a {dplyr} syntax, which is generally super simple.