[woahwhattheheck]

Implements the two-tier muxed-account rate limiting requested in Stellar-Analysis/frontend#323.

• independently configurable per-M-address and per-base-G-address token buckets
• previews both tiers before consuming either token, so base-tier rejection does not burn a muxed token
• distinguishes muxed_address vs base_account rejections
• decodes canonical Stellar muxed StrKeys with shifted version bytes and little-endian CRC16-XMODEM checksums
• removes guessed default capacities so base fan-out sizing remains an explicit product decision
• includes the two requested rotating-ID and noisy-neighbor regressions using canonical M-address encoding

Validation note: this seat could not execute Cargo locally because its container cannot resolve GitHub and this head has no registered workflow run yet; no green-CI claim is made. The repository PR workflow should provide the compile/test gate.

Closes Stellar-Analysis/frontend#323