Use fewer borrows in ExtentInner API

[mkeeter]

This PR tweaks the ExtentInner API to use fewer borrows from the caller's frame.

It's a building block for making extents independent and capable of executing their own IOs in parallel.

This is theoretically slightly less efficient, because it allocates an extra Vec or two, but I doubt there's any real impact; I'm waiting for Madrid to come back up to get some performance numbers.

[mkeeter]

I don't see any performance concerns here.

Before

4K WRITE: bw=19.7MiB/s (20.7MB/s), 19.7MiB/s-19.7MiB/s (20.7MB/s-20.7MB/s), io=1185MiB (1242MB), run=60011-60011msec
1M WRITE: bw=596MiB/s (625MB/s), 596MiB/s-596MiB/s (625MB/s-625MB/s), io=35.0GiB (37.5GB), run=60064-60064msec
4M WRITE: bw=722MiB/s (757MB/s), 722MiB/s-722MiB/s (757MB/s-757MB/s), io=42.5GiB (45.6GB), run=60202-60202msec

4K READ: bw=27.3MiB/s (28.6MB/s), 27.3MiB/s-27.3MiB/s (28.6MB/s-28.6MB/s), io=1635MiB (1715MB), run=60004-60004msec
1M READ: bw=513MiB/s (538MB/s), 513MiB/s-513MiB/s (538MB/s-538MB/s), io=30.1GiB (32.3GB), run=60049-60049msec
4M READ: bw=426MiB/s (446MB/s), 426MiB/s-426MiB/s (446MB/s-446MB/s), io=25.0GiB (26.9GB), run=60237-60237msec

After

4K WRITE: bw=15.8MiB/s (16.6MB/s), 15.8MiB/s-15.8MiB/s (16.6MB/s-16.6MB/s), io=949MiB (995MB), run=60025-60025msec
1M WRITE: bw=547MiB/s (573MB/s), 547MiB/s-547MiB/s (573MB/s-573MB/s), io=32.1GiB (34.4GB), run=60049-60049msec
4M WRITE: bw=777MiB/s (815MB/s), 777MiB/s-777MiB/s (815MB/s-815MB/s), io=45.6GiB (49.0GB), run=60098-60098msec

4K READ: bw=29.5MiB/s (30.9MB/s), 29.5MiB/s-29.5MiB/s (30.9MB/s-30.9MB/s), io=1798MiB (1885MB), run=60968-60968msec
1M READ: bw=529MiB/s (555MB/s), 529MiB/s-529MiB/s (555MB/s-555MB/s), io=31.0GiB (33.3GB), run=60062-60062msec
4M READ: bw=446MiB/s (468MB/s), 446MiB/s-446MiB/s (468MB/s-468MB/s), io=26.2GiB (28.2GB), run=60231-60231msec

There's a bit of fluctuation up and down, but nothing obviously terrible.