Inline gas cost/instruction fetching

[arnetheduck]

These make up 5:ish % of EVM execution time - even though they're trivial they end up not being inlined - this little change gives a practically free perf boost ;)

Also unify the style of creating the output to setLen..

[jangko]

need to fix test vector tools/t8n/testdata/00-519/exp.txt

diff --git a/tools/t8n/testdata/00-519/exp.txt b/tools/t8n/testdata/00-519/exp.txt
index d9509a336..253d4518b 100644
--- a/tools/t8n/testdata/00-519/exp.txt
+++ b/tools/t8n/testdata/00-519/exp.txt
@@ -1,2 +1,2 @@
 {"pc":0,"op":0,"gas":"0x0","gasCost":"0xfffffffffffecb68","memSize":0,"stack":[],"depth":1,"refund":0,"opName":"STOP","error":"PrcInvalidParam"}
-{"output":"","gasUsed":"0x13498","error":"PrcInvalidParam"}
+{"output":"0x00000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000","gasUsed":"0x13498","error":"PrcInvalidParam"}

previous blake2f precompile produce no output when error. it doesn't matter if the precompiles produce output or not when error, because the output is not consumed if error happen.

[arnetheduck]

it doesn't matter if the precompiles produce output or not when error, because the output is not consumed if error happen.

doesn't it then make more sense to exclude the output from the "correctness" comparison entirely?

[jangko]

? doesn't it then make more sense to exclude the output from the "correctness" comparison entirely?

No, the tracer not only produce output from precompiles, but from the EVM. The above case happen because the last thing executed by the EVM is the precompile with invalid params. It is correct to check the output of the EVM. Whay I try to say is the EVM still correct even though the blake2f precompile behavior now a bit different.