refactor: centralize SCALE codec primitives in utils/scale.py

[fansilas]

Summary

SCALE primitive encoders and decoders were duplicated across three modules: contract_client.py, validator/event_watcher.py, and validator/axon_handlers.py.

Each copy carried its own magic byte widths, hex-prefix stripping, and compact-length encoders. This collapses them into a single allways/utils/scale.py that all three modules now import from.

No behavior change for well-formed contract or RPC data. Where the three copies disagreed on error handling (out-of-bounds decode_string, mode-2 compact address length, stray 0x substrings), the extracted module picks the more defensive variant so malformed input fails gracefully rather than silently corrupting output.

Changes

New module: allways/utils/scale.py

• Constants: SS58_PREFIX, U32_BYTES, U64_BYTES, U128_BYTES, ACCOUNT_ID_BYTES
• Helpers: strip_hex_prefix, compact_encode_len, encode_bytes, encode_str, encode_u128
• Streaming decoders returning (value, new_offset): decode_u32, decode_u64, decode_u128, decode_bool, decode_account_id, decode_string

allways/contract_client.py

• Dropped the local compact_encode_len definition and inline hex-prefix / u128 / byte-width logic
• encode_value, extract_*, decode_swap_data, get_reservation_data, and raw_contract_read now delegate to the shared helpers
• Removed the one-line decode_string method that just forwarded to the utility
• Byte-width constants replace magic 4, 8, 16, 32 literals

allways/validator/event_watcher.py

• Dropped the parallel copies of decode_u32/u64/u128/bool/account_id/string in favor of imports from utils/scale
• topic_account_id delegates to decode_account_id, and the hex-prefix strip in to_bytes uses strip_hex_prefix
• struct and ss58_encode imports removed (no longer needed locally)

allways/validator/axon_handlers.py

• The three scale_encode_*_hash_input functions now compose encode_bytes, encode_str, and encode_u128 instead of inlining compact_encode_len(len(...)) + ... and .to_bytes(16, 'little')
• The local encode_str lambda inside scale_encode_initiate_hash_input is gone

tests/test_scale.py

• TestDecodeString exercises decode_string from the utility directly now that the client shim is gone
• compact_encode_len imported from its new home

Testing

• All 267 tests pass (pytest)
• ruff check allways/ tests/ clean
• ruff format --check allways/ tests/ clean
• u128 encoding verified bit-identical against struct.pack('<QQ', lo, hi) across the range [0, 2**128 - 1] including boundaries at 2**64 and 2**127
• String round-trip passes in all three SCALE compact-length modes (lengths 0, 5, 64, 16383, 16384)
• SS58 encoding via ss58_encode(bytes, 42) produces the same output as the previous substrate.ss58_encode(bytes.hex()) path on Bittensor

[LandynDev]

Hi @fansilas , thanks for the contribution. I think this could make the codebase cleaner from a quick skim, can you fix conflicts before I look more into it?

[fansilas]

Hi @LandynDev
Thanks for your review. I've fixed the conflict. Can you please review again?