fix(marshal): Avoid bigint literals for Google Apps Script compatibility

[gibson042]

Ref #1582

This doesn't fix the redefined issue, but hopefully does remove the incompatibility that originally prompted it.

[dckc]

Would this change be visible in XS heap snapshots? I hesitate to approve until we're quite clear on how we're dealing with endo/agoric-sdk compatibility going forward.

[mhofman]

Would this change be visible in XS heap snapshots? I hesitate to approve until we're quite clear on how we're dealing with endo/agoric-sdk compatibility going forward.

Changes in Endo will only be reflected once the vat is upgraded, either the lockdown / supervisor bundle and / or the contract itself. I think it's safe to merge changes in Endo, unless they are dealing with some kind of protocol that isn't backwards compatible (e.g. a new marshal encoding/serialization)

[dckc]

Changes in Endo will only be reflected once the vat is upgraded, either the lockdown / supervisor bundle and / or the contract itself.

Right... but suppose we decide later that we want to make a change in endo... call it B... one that does not impact the supervisor bundle. But we had already merged this change--call it A--which does impact the supervisor bundle. And we want to start new contracts/vats that take advantage of B. Are we forced to start them with a supervisor bundle that includes A? How do we even do that? Suppose it's not automatic, but we forget/neglect to do it... does the vat get launched with the pre-A supervisor bundle? Is it compatible?

I'm not going to stand in the way / request changes. But Google Apps Script compatibility is not a clear product requirement. Until we have changes motivated by a clear product requirement and we have a clear story on how they're going to be deployed, I'm not inclined to approve.

[erights]

FWIW the change looks good to me. But I'm not approving to leave @dckc 's concern in place.

Separately, do we have any idea what other impediments there are to using marshal in Google Sheets?

It seems weird to me that there is a version of JS that does have bigints and bigint arithmetic, but does not parse bigint literals. How confident are we that this parsing issue is a limitation of the JS engine and not some earlier tool in their internal toolchain?