AMR.load and AMR.amr_to_string changes data during read in test/amr_examples

amr_to_string appears to change data as its being read. This means that a test case `(AMR.amr_to_string(AMR.load(ASKEModel, model_expr))) == modelrep` will always fail. Here are two concrete observations of this issue:

1. "@" character is omitted on `AMR.load`
**Expected** Call the argument of `amr_to_string` in the test case `M`. A substring of `M.header` should read:
"...ASKE Model Representation: SIR Model@0.1..."
**Observed** The substring of `M.header` omits the "@" character:
"...ASKE Model Representation: SIR Model0.1..."
2. A match case in `amr_to_string` erroneously interprets `ODERecord` as `ODEList`
**Expected** The `ODERecord` is interpreted by the `ODERecord` match case and return "ODE_Record = ..."
**Observed** The relevant substring in the output of `amr_to_string` has "ODE Equations" instead.