add unit test for find_secondary host functions

[spoonincode]

Which key should the find_secondary host function return when there are duplicate keys? cdt avoids this question entirely by never using these host functions and instead always doing a lowerbound (it's possible the find_secondary host function has never once been used on EOS! Unless a rare contract built with something other then cdt used it of course)

The host function ultimately performs a "partial key lookup" on the table (this then goes through to chainbase's find())

spring/libraries/chain/include/eosio/chain/apply_context.hpp

Lines 291 to 297 in f871bf2

	int find_secondary( uint64_t code, uint64_t scope, uint64_t table, secondary_key_proxy_const_type secondary, uint64_t& primary ) {
	auto tab = context.find_table( name(code), name(scope), name(table) );
	if( !tab ) return -1;
	auto table_end_itr = itr_cache.cache_table( *tab );
	const auto* obj = context.db.find<ObjectType, by_secondary>( secondary_key_helper_t::create_tuple( *tab, secondary ) );

In leap 3.1+ & spring, the underlying data structure (boost intrusive avltree) does specify that find() is treated like lowerbound(). And even looking through the code makes it clear that find() simply calls lowerbound().

EOSIO 2.0 and earlier it is not so clear: the documentation for the data structure in use (boost multindex directly) simply says it finds "a" matching key. But eyeballing the code suggests it does effectively behave like lowerbound. And empirically the test code still passes on 2.0.

This test confirms that the find_secondary host functions behave like lowerbound: lowest primary key is returned.

Closes #1019

[spoonincode]

hm maybe I should have just done this in a .cpp contract, but I would have had to use the internal_use_do_not_use functions which I wanted to avoid.

[ericpassmore]

Note:start
category: Tests
component: Host Functions
summary: Add unit test for find_secondary host functions.
Note:end