Add notes for vl=0 behavior of instructions that write scalars

Resolves #703

* vstart needs explicit initialization by software at reset

* Simpler fix that preserves usefulness of original commentary

First release candidate for v1.0 for public review.

defining prestart, inactive, active, body, and tail sections, and
corrected discussion of what elements might be updated.
Closes #689.

Add space between VType's operands

Closes #687.

vm=0 encodings should be reserved for mask-logical instructions

Changed assembler mnemonic of unordered floating-point reductions `vfredsum` and `vfwredsum` to `vfredusum` and `vfwredusum` respectively.  Older names kept as aliases.
Closes #681.

Clarified that mask load and store instructions operate as if EEW=8, EMUL=1 with reduced vector length.
Closes #665.

Closes #652.

Allow mask-agnostic tails to be written with the value that would have been produced with a maximum vector length.
Closes #639.

Closes #640.

Closes #612.

Made clear that implementations can raise illegal instruction exceptions if EEW not supported for address offset in indexed vector memory instructions.
Closes #599

Closes #587.

Removed the "write 0" request for vtype.  Can only be written via  vsetvl and that definition gives correct behavior.
Closes #575.

Will be added as a separate vector extension at a later date but with a new encoding.
Closes #545.

and moved to standard extension section at end.

Closes #550.

Clarify that whole-register loads do use the encoded EEW

The old text points out that, because the in-memory layout and in-register
layouts are the same, the instructions behave the same as moving vectors
of bytes with EEW=8.  True enough as far as the data transfer, but not so
for the handling of evl/vstart: those actually do depend on the encoded EEW.

This choice is consistent with the later statement that "implementations are
allowed to raise a misaligned address exception on whole register loads",
meaning (a) that EEW is not truly a HINT and (b) the unit of transfer really
is elements, not bytes, even if the effect of a successful transfer is the
same.

Resolves #686

The value 0x1000 doesn't fit in 3 bits. I assume this is supposed to be 0 since the other formats use 1-6?

Resolves #679

Fix incorrect statement about vasub overflow