fix(codec): Fix flat encoding and decoding of arbitrarily size integers (#378)

This commits fixes the flat encoding and decoding (and consequently,
  the zigzag) for large integers in the following ways:

  - It removes support for encoding and decoding i128 values.

  - It optionally (feature = "num-bigint") introduces encoding and
    decoding of large sized integers through the num-bigint::BigInt
    type.

  Without the feature enabled, it is still possible to encode and decode
  isize values; but the use of i128 is now prohibited (as it would
  overflow on boundaries) in favor of arbitrarily sized integers.

  The commit also introduces a missing property roundtrip for encoding
  and decoding large integers, which was missing and thus, failed to
  identify the overflow problem.

  See related issue: https://github.com/aiken-lang/aiken/issues/796

pallas-codec/src/flat/decode/mod.rs

@@ -3,6 +3,9 @@ mod error;
 
 use crate::flat::filler::Filler;
 
+#[cfg(feature = "num-bigint")]
+use num_bigint::BigInt;
+
 pub use decoder::Decoder;
 pub use error::Error;
 
@@ -36,9 +39,10 @@ impl Decode<'_> for isize {
     }
 }
 
-impl Decode<'_> for i128 {
+#[cfg(feature = "num-bigint")]
+impl Decode<'_> for BigInt {
     fn decode(d: &mut Decoder) -> Result<Self, Error> {
-        d.big_integer()
+        Ok(d.big_integer()?.into())
     }
 }