Enhancement: Refactor handling of index files and Inventory

[blaylockbk]

The goal of this PR is to refactor the handling of the GRIB index files. This introduces a new Inventory class that is used to check if the index file exits, read the index file into a Pandas DataFrame, and filter the inventory using a searchString which is used for downloading file subsets. The only input to this class is the local or remote path to the index file.

from herbie.inventory import Inventory

I = Inventory("https://noaa-hrrr-bdp-pds.s3.amazonaws.com/hrrr.20231201/conus/hrrr.t00z.wrfsfcf00.grib2.idx")

# All the data in the index file
df = I.dataframe

# A subset of the GRIB messages
df_subset = I.filter("UGRD|VGRD")

This should make working with index files a little more maintainable and debugable.

• Write sufficient tests.
• Test for all model types.
• Implement in the Herbie class.