[CDB-READER] Optimize `BeamData`

[StudioWEngineers]

load() optimizations:

• avoid new Dataframe at each iteration
• avoid isin when assigning groups

_load() optimizations:

• use bool instead of int for position indicator
• change the match with dict lookup

get() optimizations:

• use multindex