Update polyfill.js for browsers with poor support

[brettz9]

Although this issue may tie in to all of the browser-specific (Safari) issues (hopefully all resolvable for Milestone 3.0.0 (tracking in #262)), I think it should be helpful for us to have a specific issue for tracking the updating of polyfill.js, as that file for browsers with poor support like Safari IE has fallen majorly behind compared to the complete browser shim (and Node), including for newly added APIs like getAllKeys and such.

[brettz9]

For those not aware, the original maintainers of the project have not been able to be very active of late. I came on board with an interest in helping bring the Node implementation up to speed.

This could have potential gains for the browser (and it should in Chrome and Safari 10+ if using the shim, though the shim is less necessary and still in need of having those theoretically more surmountable test failures fixed that are not failing in the Node implementation).

However, as my interest and focus has not been on iOS9 and non-Chrome Android IE which the original code sought to selectively patch given partial but poor IndexeDB support, the improvements (which were part of a file, polyfill.js have not been kept much up to date and may in fact now be broken.

Either we need:

1. volunteers who will try to apply the new improvements to this old polyfill (or at least ensure there are no regressions). Depending on whether continuing problems are surmountable, we may also need to apply this piecemeal patching for Safari 10 later IE versions.
2. to remove polyfill.js entirely.

I know the latter is not very satisfying, especially since the largest base of users was using this for browsing polyfilling (the Node.js support was only added to master and has not yet been added to an npm release). However, I feel that with a new major (i.e., breaking) version, at least making the Node support available ought to be worthy of a release. But we should be clear on where we are leaving things with the various browsers.

[brettz9]

Ok, so the plan is now to abandon polyfill.js for [3.0.0] (#262) (unless someone testing in (older?) IE wishes to salvage it and keep it up to date).