Comm host client proxy return rollup blob

[bxue-l2]

Alternative approach to integrate with eigenda to #18

This PR does the follows

• host fetch blob from proxy, assuming the proxy returns the original rollup-data (i.e compressed batches)
• host does the codec, and prepend the blob_header (containting the rollup-date.len())
• host splits data into field element by field element, where key is hash of 96 bytes. Where the first 64 bytes contain kzg commitment, the last 8 bytes contains the field element index
• host does not provide a proof at the end
• client can read field element by field element
• client check if any returned byte is empty, return early if true
• once the entire blob is received, the eigenda_data.rs removes the header again. Note we don't do IFFT/FFT anywhere
• Inside the eigenda.rs, we perform a check on blob staleness. However, it is not a complete feature.

This PR contains #17

[samlaf]

is there a reason why we make this 96 long instead of 72? [65,88) is always just 0 bytes.

[bxue-l2]

curerntly, op uses 80 Bytes, even though their commitment is only 48 bytes. there is extra 32 Bytes.
Since our commitment is 64Bytes, so in total 96.

However, I recall the extra 32 bytes for op 4844 is for bytes alighment. We can change it later

[samlaf]

@bxue-l2 I just rebased + fixed the README merge conflict.

[bxue-l2]

We should pull the decode function from eigenda_data.rs into a separate module that handles both encode and decode, called it codec.

So that the bin/host/src/eigenda_fetcher/mod.rs can call the function. Inside the codec, we can more robustly do the test