Replace furrr with mirai for parallel execution

[NewGraphEnvironment]

Replace furrr with mirai for parallel execution

Problem

furrr/future creates heavyweight worker processes per future_map() call. Workers are ephemeral — no connection reuse, serialization overhead. Each worker must call frs_db_conn() independently and relies on PG_*_SHARE env vars.

Proposed Solution

Switch to mirai for parallel map in frs_habitat(). mirai provides:

• Lightweight persistent daemons via NNG (nanomsg)
• mirai_map() for parallel map
• Lower overhead than future/furrr
• Foundation for crew.aws.batch (crew.aws.batch for province-wide parallel pipeline #76) — crew is built on mirai

Scope

• Replace ~20 lines of furrr/future code in R/frs_habitat.R
• Update DESCRIPTION: remove furrr/future from Suggests, add mirai
• Pass DB connection params to workers (not env-var dependent)
• Two parallel sections: Phase 1 (partition prep) and Phase 2 (species classification)

Relates to

• crew.aws.batch for province-wide parallel pipeline #76 — crew.aws.batch (mirai is prerequisite)
• Profile and optimize Phase 2 species classification performance #72 — performance optimization