signer: create low memory feature for incoming vls bytes

[irriden]

pull vls chunked-filter work, see chunked-buffer, serde_bolt, txoo crates for details.

The tldr is that we now use the chunked_buffer::GenericChunkedBuffer structure, imported here as serde_bolt::NonContiguousOctets for both the MQTT bytes, and the ProofType::Filter structure in txoo.

Key features of NonContiguousBytes:

• Max contiguous memory block size: 1024 bytes.
• Bytes are progressively de-allocated from memory as they are read.
• As bytes are written to the structure, no reallocation of the bytes themselves ever occurs. We instead keep allocating independent, 1024-byte blocks of memory as necessary.

The use of this structure achieves the following:

• Completely eliminates the source of all of our out-of-memory crashes up until this point.
• Reduces the max contiguous memory block size from 45KB to 1KiB when creating both of these structures, which increases the efficiency of our memory usage.
• Decreases peak memory usage by progressively deallocating the mqtt bytes and allocating the ProofType::Filter bytes as the parse occurs. We previously needed space for 2 45KB contiguous memory blocks, (1 for the MQTT bytes, and one for the Filter bytes). This resulted in very premature out-of-memory crashes, ie memory crash when still 90KB of memory available. This is because the MQTT bytes were deallocated only once the full Filter structure had been created.

total: 91K, max block: 37K
msgs::read
memory allocation of 39235 bytes failed

[Evanfeenstra]

awesome work on this @irriden !