[breaking/format] Remove vlen from entry header

[jarifibrahim]

This commit removes value length from the entry header stored in each SST.
We don't need vlen to store the length of the value. We can find it by
using the entry offsets in the footer of the blocks.

Entries in the table are of the form

Klen1 (Point A)	Plen1	Key1	Value1	Klen2 (Point B)

And we have the entry index at the end of each block.

Entry 1 offset (Point A)	Entry 2 Offset (Point B)

Using the entry index and current position in the buffer, we can find
the length of the value.

Note this PR also reduces the size of the header from 8 bytes to 4 bytes.

---

This change is 

[manishrjain]

Test it thoroughly. Great change, but just ensure that nothing is breaking.

Reviewable status: 0 of 3 files reviewed, 2 unresolved discussions (waiting on @jarifibrahim and @manishrjain)

---

table/iterator.go, line 46 at r1 (raw file):

	itr.pos = 0
	itr.err = nil
	itr.baseKey = itr.baseKey[:0]

Not so sure about this change. Have to be sure that we're not using that array anywhere else.

---

table/iterator.go, line 165 at r1 (raw file):

	// We're at the last entry in the block.
	if itr.currentIdx == itr.numEntries-1 {
		valEndOffset = uint32(itr.entriesIndexStart)

Is it possible that this might be zero?

[manishrjain]

Reviewable status: 0 of 3 files reviewed, 3 unresolved discussions (waiting on @jarifibrahim and @manishrjain)

---

table/iterator.go, line 176 at r1 (raw file):

		return
	}
	itr.val = y.SafeCopy(itr.val, itr.data[itr.pos:valEndOffset])

Add a TODO here to check if we can avoid this copy.

[jarifibrahim]

Reviewable status: 0 of 3 files reviewed, 3 unresolved discussions (waiting on @manishrjain)

---

table/iterator.go, line 46 at r1 (raw file):

Previously, manishrjain (Manish R Jain) wrote…

Not so sure about this change. Have to be sure that we're not using that array anywhere else.

We do a Copy(..., itr.basekey) everywhere.

---

table/iterator.go, line 165 at r1 (raw file):

Previously, manishrjain (Manish R Jain) wrote…

Is it possible that this might be zero?

No. Every time table.Finish() would be called we will write all the entries to the end of the block and entriesIndexStart will point to that offset.

---

table/iterator.go, line 176 at r1 (raw file):

Previously, manishrjain (Manish R Jain) wrote…

Add a TODO here to check if we can avoid this copy.

Done.

[manishrjain]

Reviewed 2 of 3 files at r1, 1 of 1 files at r2.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @manishrjain)